INFORMATICS ENGINEERING JOURNALS & RESEARCH
Transcript of INFORMATICS ENGINEERING JOURNALS & RESEARCH
Volume 4 | Number 2 | April 2020
E-ISSN : 2541-2019
P-ISSN : 2541-044X
INFORMATICS ENGINEERING JOURNALS & RESEARCH
Dipublikasi oleh :
PREFACE
Praise the presence of Allah SWT, for blessings and mercy, in the successful
publication of the Journal SinkrOn Volume 4 Number 2 for the April 2020 Period which
refers to the rule for journal writing determined by Polytechnic Ganesha Medan.
SinkrOn as a media publication and a place to share scientific work in the field of
Informatics Engineering which is published twice in a year, namely in October and April,
and this edition is filled with contribution from a various institution in parts of Indonesia
with research fields related to the field of Informatic Engineering from lectures and graduate
students.
This edition and so on SinkrOn will use international language, namely English, and
will progressively go to International Journal in hopes of reaching out to writers and readers
from all over the world. In this edition, SinkrOn also involving several editors and reviewers
from various countries such as Indonesia, Malaysia, India, Iraq, and Italy.
Our deepest gratitude to the authors, reviewers, editors, and all parties involved in
publishing this journal. Hopefully, this journal can provide good benefits for all academics
in the field of Informatics Engineering and still waiting for your brilliant work in the next
edition.
Medan, 6 April 2020
Editor in Chief
Table of Contents
Preface
Table of Contents
Sinta Accreditation certificate
Application of Pizza Sales Data Mining Using Apriori Method
Rusdiansyah, Nining suharyanti, Triningsih, Murniyati
1-5
Classification Of Borax Content In Tomato Sauce Through Images Using
GLCM
Reyhan Achmad Rizal, Mario Susanto, Andy Chandra
6-9
Application of Expert System for Diagnosing Diseases Cocoa Plants Using the
Forward Chaining Algorithm Method
Omar Pahlevi, Muhamad Kusumo Atmojo
10-18
Analysis of Multi-attribute Utility Theory for College Ranking Decision Making
Adidtya Perdana, Arief Budiman
19-26
Decision Support System For Determining the Best College High Private Using
Topsis Method
Yuyun Dwi Lestari, Mardiana Mardiana
27-33
Pears Classification Using Principal Component Analysis and K-Nearest
Neighbor
Moh. Arie Hasan, Arief Setya Budi
34-41
Analysis K-Nearest Neighbor Algorithm for Improving Prediction Student
Graduation Time
Rizki Muliono, Juanda Hakim Lubis, Nurul Khairina
42-46
Apriori Algorithm On Car Rental Analysis With The Most Popular Brands
Leo Fernando Panjaitan, Yopi Handrianto, Achmad Nurhadi
47-55
Selection of Outstanding Lecturers with Simple Additive Weighting Method
Embun Fajar Wati, Istikharoh Istikharoh, Tuslaela Tuslaela
62-67
Expert System Detects Problems of Inkmaker Machine And Mixer With
Forward Chaining
M. Sinta Nurhayati, Rachmat Hidayat
68-75
The Use of Apriori Algorithm in the Formation of Association Rule at Lotteria
Cibubur
Ovi Liansyah, Henny Destiana
76-84
Decision Support System for Achieving Scholarship Selection by Using Profile
Matching Method
Rani Irma Handayani, Triningsih, Melia Putri
92-97
Dog Disease Expert System Using Certainty Factor Method
Linda Marlinda, Widiyawati Widiyawati, Wahyu Indrarti, Reni Widiastuti
98-104
Support Vector Machine Parameter Optimization to Improve Liver Disease
Estimation with Genetic Algorithm
Hani Harafani
106-114
Simple Additive Weighting for Decision Support Selection of Expedition
Services
Suhar Janti, Mohammad Adriansyah, Ghofar Taufik
115-122
Pythagoras Tree Applied For Determined Instagram Usage Habit Decision
Erlin Windia Ambarsari, Herlinda
56-61
Analysis of Taklinear Performance and Integer Linear Programming Models in
Nurses Scheduling Problems
Junerdi Nababan, Tulus Tulus, Zakarias Situmorang
123-129
Evaluation of the Satisfaction of Users of Weather Forecast Systems with the
Service Quality Method
Eva Rianti, Syafrika Deni Rizky, Fariz Haris Nugraha
130-140
Analysis of Cadets Satisfaction to Medan Aviation Polytechnic Service Using
Quality Function Deployment (QFD) Method
Liber Tommy Hutabarat, Lisda Juliana Pangaribuan
141-150
The use of M-Learning in Teaching English for Civil Engineering Students
Dharmawati
151-155
SMART and TOPSIS Method For Determining The Priority Of Screen Printing
Rinaldi Setiawan, Arini Arini, Luh Kesuma Wardhani
151-157
Deep Neural Networks Approach for Monitoring Vehicles on the Highway
Amir Mahmud Husein, Christopher Christopher, Andy Gracia, Rio Brandlee,
Muhammad Haris Hasibuan
163-171
Optimizing Genetic Algorithms for Sentiment Analysis of Apple Product
Reviews Using SVM
Elly Indrayuni, Acmad Nurhadi
172-178
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10500
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative
Commons Attribution-NonCommercial 4.0 International License.. 1
Application of Pizza Sales Data Mining Using
Apriori Method
Rusdiansyah
Bina Sarana Informatika University
West Jakarta Indonesia
Nining Suharyanti
Bina Sarana Informatika University
West Jakarta City, Indonesia
Triningsih
Bina Sarana Informatika University
West Jakarta Indonesia
Murniyati
Bina Sarana Informatika University
West Jakarta Indonesia
Submitted: Feb 19, 2020
Accepted: Mar 27, 2020
Published: Apr 1, 2020
Abstract— Pizza is a processed food originating from Italy and has been spread in
various other countries including one of them in Indonesia. Pizza is a processed food
that is currently sought after by various groups of people so as to make the pizza
business opportunity very profitable, if it is run in a food business. Currently the pizza
business has very favorable prospects when compared to other businesses. Moreover,
the targeted target can be from all walks of life from children to adults. Pizza sales
transactions that produce sales data every day, have not been able to maximize the use
of sales data. Sales data is only stored as an archive, so it becomes a pile of data.
Therefore the use of data mining is used to solve this problem. A priori algorithm is a
data mining method by using minimum support parameters, minimum confidence and
will analyze in the period of every month of sales transactions. This study produces
data on the results of the process of association rules from the data collection of sales
transactions. From the association rules it can be concluded that the pattern of pizza
sales, where consumers more often buy Meatzza and Cheese Mania, as evidenced by
the results of calculations using Apriori Algorithm and Rapidminer 5.3, with support
of 30% and 60% confidence.
Keywords— Pizza Sales, Apriori Algorithms, Association Rule
I. INTRODUCTION
Pizza is a processed food that is currently sought after by various groups of people, thus making the pizza business opportunity very profitable, if run in the food business. To start a business selling Pizza, including one selling fast food. (Adibah & Novita, 2019). The Pizza sales business has quite a number of competitors, so skills are needed to attract the attention of consumers by refining, perfecting, and providing various pizza variants. Business activities are marked by intensifying competition between one business actor and another business actor.
Pizza is quite popular among the people and
quite popular from children to adults, research to
examine which pizza is most often bought by
consumers. To find out the buyer's interest, one of
the information that can be obtained is from the
Sales Data, from the Sales Data it will show which
pizza variants are often bought by consumers.
Problems that arise can be identified as follows:
1. A lot of data that has not been neatly organized
and only stored without being processed into
information that is useful for increasing sales.
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10500
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License.. 2
2. Choosing a pizza variant that is too much for
new consumers to have difficulty in choosing,
causing the queue to be too long and take longer
and occasionally errors in consumer orders.
II. LITERATURE REVIEW
Simply put Data Mining is mining or discovering
new information by looking for certain patterns or
rules from a very large amount of data (Rismayanti,
Damayanti, & Khairunnisa, 2019), Data mining is a
method for finding knowledge in a large pile of data.
Data mining is the process of digging and analyzing
a very large amount of data to obtain something that
is true, new and useful and finally can be found a
pattern in the data (Rodiyansyah, 2015). A priori
algorithms include types of association rules in Data
Mining. Rules that state the association between
several attributes are often called affinity analysis or
market basket analysis. Association analysis or
association rule mining is a Data Mining technique
to find the rules for a combination of items. One of
the stages of association analysis that attracts many
researchers to produce efficient algorithms is
analysis of high frequency patterns (frequent pattern
mining) (Badrul, 2016). FP-Growth is a
development of a priori algorithm. A priori
algorithm requires generating candidates to get
frequent itemsets (Meilani & Azinar, 2015).
However, in the FP-Growth algorithm, generating
candidates is not done because FP-Growth uses the
concept of tree development called Frequent Pattern-
Tree (FP-Tree) in the search for frequent itemsets.
That causes FP-Growth algorithm to be faster than
Apriori algorithm. So the shortcomings of the a
priori algorithm are corrected by the FP-Growth
algorithm. The FP-growth algorithm can be
implemented in a short time and produces high
accuracy when used to analyze the association rule
(Muliono, 2017).
III. PROPOSED METHOD
The research instrument is a tool or facility used
by researchers in gathering data to make work easier
and the results better, in the sense of being more
accurate, complete, and systematic so that it is easier
to process (Nasution, 2019).
The author uses quantitative data analysis
in research taken by the author, this research which
is an analysis prioritizing data in the form of
numbers and also calculations using formulas related
to writing analysis, in this case Apriori Algorithm
analysis will be used as follows:
A. Analysis of sales problems Problems will be
analyzed using the Apriori Algorithm Method.
B. Data processing with a priori algorithm
calculation
The following are some of the stages that will
be performed in calculations with the Apriori
Algorithm (Satie, Suparni, & Pohan, 2020):
1. Look for the biggest value that has sold the
most
2. The initial step is to find the highest sales value
in a month's transaction data with the following
steps:
a. determine the pizza list
b. determine pizza sales data
3. Determine pizza sales data
4. Grouping some of the most frequently
purchased pizza.
5. Making a Tabular Format
6. If the biggest sales value from the sale of pizza
that is often bought every month is known, it
will be made tabular format so that it can be
analyzed with Algorima Apriori.
High Frequency Pattern Analysis in this stage to find
a combination of items that meet the minimum
requirements of the support value in the database,
which can be formulated as follows (Wahyuningtias
& Rusdiansyah, 2019):
𝑆𝑢𝑝𝑝𝑜𝑟𝑡(A) = Number of Transactions 𝐴 𝑥 100%
Transaction Total
Meanwhile, the formula 2 itemset with the formula:
𝑆𝑢𝑝𝑝𝑜𝑟𝑡(A, B) =Number of Transactions A and B𝑥 100%
Transaction Total
The search for high frequency patterns will be
stopped if the combination does not meet the
specified Support value.
IV. RESULT AND DISCUSSION
Transaction data that occurs every day to make it
easier with a large amount of data because of sales
activities every day the data is getting more and
more time. The place has not been said to be neatly
arranged in the processing of these data only stored
and archived not utilized and processed into useful
information to increase sales.
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10500
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License.. 3
TABLE I. LIST OF PIZZA VARIAN PRODUCTS
PIZZA DATA TABLE
Number PIZZA CATEGORY (MEDIUM,
THIN CRUST)
1 CM CHEESE MANIA
2 PF PEPPER FEAST
3 CL CHK LOVERS
4 ACCB
AMR CLS
CHEESEBURGER
5 MT MEATZZA
6 EX EXTRAVAGANZZA
7 AS AMR ALL STAR
8 GBS
GRILLED BEEF
SUPREME
9 DBB DBL BEEF BRGR
10 CHS CHEESY SAUSAGE
11 VM MEAT n MEAT
12 TNDE TUNA DELIGHT
13 CDO CHK DOMINATOR
14 BDL BEEF DELIGHT
15 CPF CHK PEP FEAST
16 MAR MARGHERITA
17 CHD CHICKEN DELIGHT
18 SBB SAMBAL BEEF
Item set formation
Settlement based on Table 1 provided by the
process of forming C1 or referred to as Itemset,
Itemset with a Minimum Amount of 30%.
From the process of forming an itemset with a
minimum support of 30%, it can be seen that those
who meet the minimum support standards are: PF,
CM, AS, TNDE, MT, ACCB, EX, VM, VEM, CL,
MAR, GBS, CDO, DBB, CPF, BDL , CHD, SBB,
CHS.Kombinasi 2 Itemset
The process of forming C2 or can be called 2
Itemset with a minimum amount of support of 30%.
Can be solved by the following formula:
From the process of forming C2 or 2 Itemset in
Table.II with a minimum support of 30%, it can be
seen that those who meet the minimum support
standards are:
TABLE II. 2 ITEM SET
Support Table of 2 Item set minimum of 30%
Itemset amount Support
S (PF , CM) 12 38.7%
S (CM , MT) 12 38.7%
S (CM , AS) 11 35.5%
S (PF , MT) 11 35.5%
Formation of Association Rules
After the High frequency pattern is found, then
look for the Association Rules that meet the
Minimum Requirements for Confidence, by
calculating the Confidence Rules Association A →
B with a Minimum Confidence of 60%. This is the
value of Rules A → B obtained.
𝐶𝑜𝑛𝑓𝑖𝑑𝑒𝑛𝑐𝑒 = (𝐴|𝐵) =Number of Transactions A and B X100
Transaction Total 𝐴
TABLE III. CONFIDENCE
1 . Confidence [CM] --> [PF]
12/18*100 =
66.7
2 . Confidence [PF] --> [CM]
12/18*100 =
66.7
3 . Confidence [CM] -->
[MT]
12/18*100 =
66.7
4 . Confidence [MT] -->
[CM] 12/15*100 = 80
5 . Confidence [CM] -->
[AS]
11/18*100 =
61.11
6 . Confidence [AS] -->
[CM]
11/18*100 =
61.11
7 . Confidence [PF] --> [MT]
11/18*100 =
61.11
8 . Confidence [MT] --> [PF]
11/15*100 =
73.33
TABLE IV. ATURAN ASOSIASI
Table Calculation of Association Rules 60%
The rules Confidence
If you buy CM, you
will buy PF 12/18*100 66.7%
If you buy PF, you will
buy CM 12/18*100 66.7%
If you buy CM, you
will buy MT 12/18*100 66.7%
If you buy MT, you
will buy CM 12/15*100 80.0%
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10500
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License.. 4
If you buy CM, you
will buy AS 11/18*100 61.1%
If you buy US, you will
buy CM 11/18*100 61.1%
If you buy PF, you will
buy MT 11/18*100 61.1%
If you buy MT, you
will buy PF 11/15*100 73.3%
System Implementation
The a priori algorithm implementation in this
study uses the Rapidminer application for testing.
Fig. 1. Model Design
In Fig 1. The design model is formed through the process steps by connecting each operator, namely: Read Excel change to Numeric to Binominal, Numeric to Binominal transfer to Fp-Growht, Fp-Growh transform to create Association Rules and Create Association Rules change to the Results Process
Then the results of the Rule are shaped into 8 rules from the results of Rapidminner 5.3 as follows:
Fig. 2. Final Association
In Fig 2. From the overall results of the a priori
algorithm implementation process in manual
calculations and calculations in Rapidminer for
pizza sales transaction data as many as 224 data by
providing a minimum limit of 30% Support and 60%
Confidence that has been set, so as to get the results
of Rule 8 as follows:
1. Pepper Feast, Meatzza with a Support Value of
0.355 and a Confidence Value of 0.6112.
2. Cheese mania, Amr All Star with a Support
Value of 0.355 and a Confidence Value of
0.611
3. Pepper Feast, Chesee mania with Support Value
of 0.387 and Confidence Value of 0.6674.
4. Cheese mania, Pepper Feast with Support Value
of 0.387 and Confidence Value of 0.667
5. Cheese mania, Meatzza with Support Value of
0.387 and Confidence Value of 0.667
6. Amr All Star, Chesee mania with a Support
Value of 0.355 and a Confidence Value of
0.688
7. Meatzza, Pepper Feast with Support Value of
0.355 and Confidence Value of 0.733
8. Meatzza, Chesee mania with Support Value of
0.387 and Confidence Value of 0.800
Fig. 3. Graph Display
In Fig 3. It can be concluded from the value of
the most superior rules with 38.7% Support and
80.0% Confidence is pizza (If you buy Meatzza then
you will buy Chesee mania).
V. CONCLUSION AND SUGGESTION
Based on the data and the results of the
discussion, the following conclusions can be drawn:
This study produces process results data from the
association rules of the sales transaction data set.
From the association's rules can be obtained by the
pattern of pizza purchases, where consumers more
often buy Meatzza and Cheese Mania, as evidenced
by the results of calculations using Apriori
Algorithms and Rapidminer 5.3, that with 30%
support and 60% confidence and the final results of
the rules 8. Support 38.7% and 80.0% Confidence
are pizza. If you buy Meatzza, you will buy Cheese
Mania.
From the description of the research the advice is
as follows:
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10500
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License.. 5
1. Data Mining with Apriori Algorithm has the
disadvantage of having to scan the database
every time, so for very large databases it takes a
long time.
In implementing the pizza sales report not only can
use a data mining method a priori algorithm
but can use other methods for future
comparison.Dalam mengimplementasikan
laporan penjualan pizza bukan hanya bisa
menggunakan data mining metode algoritma
apriori teteapi bisa menggunakan metode lain
sebagai bahan perbandingan kedepannya.
Rodiyansyah, S. F. (2015). Algoritma Apriori
untuk Analisis Keranjang Belanja pada Data
Transaksi Penjualan. Infotech, 1(1), 36–39.
Retrieved from
http://jurnal.unma.ac.id/index.php/infotech/arti
cle/view/42
2.
3. The addition of research objects does not only
cover one month but can be up to several
months backward so that data is more accurate
and in subsequent studies it is expected to
develop a computerized system based on data
mining processing using a priori algorithms to
analyze transactions that occurred in the last
period to be made business decision making
process going forward. in an application that
can be easily used for data processing.
V. REFERENCES
Adibah, F., & Novita, D. (2019). STRATEGI
UNTUK PERMINTAAN PENUH PADA
RESTAURAN M2M INDONESIAN
FAST FOOD CABANG BANGIL.
Majalah Ekonomi, 69-83. Badrul, M. (2016). ALGORITMA ASOSIASI
DENGAN ALGORITMA APRIORI
UNTUK ANALISA DATA PENJUALAN.
Jurnal Pilar Nusa Mandiri, 121-129.
Meilani, B. D., & Azinar, A. W. (2015). Penentuan
Pola Yang Sering Muncul Untuk Penerima
Kartu Jaminan Kesehatan Masyarakat
(JAMKESMAS) Menggunakan Metode
FP-Growth. Seminar Nasional “Inovasi
dalam Desain dan Teknologi” - IDeaTech
2015, 424-431.
Muliono, R. (2017). Analisis Efisiensi Algoritma
Data Mining. Semantika (Seminar Nasional
Teknik Informatika) (pp. 117-123). Medan:
Politeknik Ganesha.
Nasution, I. S. (2019). PENGARUH MODEL
PEMBELAJARAN KOOPERATIF TIPE
THINK PAIR SHARE TERHADAP
HASIL BELAJAR MAHASISWA PADA
MATA KULIAH PENGANTAR DASAR
MATEMATIKA-FKIP UMSU. MES:
Journal of Mathematics Education and
Science, 160 - 166.
Rismayanti, R., Damayanti, F., & Khairunnisa, K.
(2019). Penerapan Data Mining Algoritma
C4.5 dalam Menentukan Rekam Jejak
Kinerja Dosen STT Harapan Medan.
SinkrOn (Jurnal & Penelitian Teknik
Informatika), 99-104.
Rodiyansyah, S. F. (2015). Algoritma Apriori untuk
Analisis Keranjang Belanja pada Data
Transaksi Penjualan. Infotech Journal, 36-
39.
Satie, D. E., Suparni, S., & Pohan, A. B. (2020).
Analisa Algoritma Apriori Pada Pola
Peminjaman Buku di Perpustakaan ITB
Ahmad Dahlan. Media Informatika Budi
Dharma, 136-143.
Wahyuningtias, Y., & Rusdiansyah, R. (2019).
ANALISIS PENERAPAN ASOSIASI
UNTUK MENENTUKAN TRANSAKSI
PENJUALAN PADA WHAT’S UP CAFÉ
DENGAN METODE ALGORITMA
APRIORI. Jurnal Riset informatika, 181-
186.
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10508
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative
Commons Attribution-NonCommercial 4.0 International License. 6
Classification of Borax Content in Tomato
Sauce Through Images Using GLCM
Reyhan Achmad Rizal
University Prima Indonesia
Medan, Indonesia
Mario Susanto
University Prima Indonesia
Medan, Indonesia
Andy Chandra
University Prima Indonesia
Medan, Indonesia
Submitted: Feb 28, 2020
Accepted: Mar 15, 2020
Published: Apr 1, 2020
Abstract— One of the food products that need to be reviewed for safety and is the most
consumed is tomato sauce, although it contains a large amount of water in the sauce
which has a long shelf life because it contains acid, sugar, salt, and is often given
preservatives. The purpose of this study was to determine the tomato sauce using
harmful preservatives such as the addition of borax. The dataset used in this study is the
image of tomato sauce containing borax and not with the number of samples 400 images
of tomato sauce with different comparison percentages starting from the image of tomato
sauce with 70% borax content, image of tomato sauce with 50% borax content, image
tomatoes with 30% borax content and image of tomato sauce that does not contain borax.
A sampling of images using a camera phone brand xiaomi note 5 by mixing borax in the
original sauce before the sample is used for the training and testing process. The
classification results show the gray level co-occurrence matrix (GLCM) method is quite
optimal in classifying tomato sauce data containing borax and not with an average
percentage of the introduction of 88%.
Keywords— Tomato Sauce, Borax, GLCM
I. INTRODUCTION
The sauce is a type of flavoring that is usually added to food, the sauce can be interpreted as a thick liquid made from pasta or fruit pulp that is able to give a distinctive aroma and taste to food. Sauces can be made from fruits and vegetables such as tomatoes and chilies (Usman, Herawati, & Fitriani, 2019). Tomato sauce is made from a mixture of tomatoes and spices and then paste used is pink in accordance with the color of tomatoes used, although it contains large amounts of water the sauce has a long shelf life because it contains acids, sugar, salt, and often given preservatives (Ray, Saha, Raychaudhuri, & Chakraborty, 2016). According to (Nkhata & Ayua, 2018) a good temperature to increase the shelf life of tomato sauce is at 6 ° C. One of the food products that need to be reviewed for safety and is the most
consumed is tomato sauce, tomato sauce is usually consumed as a compliment when people consume chicken noodles, meatballs, tempura, pentol, fried rice, and others.
Many food industries are developing but not all are honest in their processing, such as the addition of harmful preservatives or the provision of preservatives that are not in accordance with the recommended quantities such as adding borax to the sauce to increase the shelf life of the sauce, this can cause health problems for those who consume it. Borax is a white crystalline powder containing boron, borax is generally used for anti-fungi, wood preservatives, and antiseptic ingredients. Consumption of foods containing borax can cause poisoning and can even cause death (Lathifah, Turista, Azizah, & Khulaifi, 2019). Making sauces
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10508
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License. 7
that are safe and healthy without the addition of preservatives can be used as food products that are good for health or can also be done to adjust the pH and temperature power in the production process. Setting the pH and temperature in the process of making sauces can improve the nutritional value, taste and long-term shelf life (Nursari, Karimuna, & Tamrin, 2016).
Based on these problems, research is needed to determine which tomato sauce is feasible and not suitable for consumption. In this study a tomato sauce classification system will be made through an image using the gray level co-occurrence matrix (GLCM) with sample data generated by researchers through a cellphone camera, where Gray Level Co-Occurrence Matrix (GLCM) method has quite good stages in the process of object recognition, as for the stages, namely: the initial stages of forming a matrix of two-pixel pairs that line in the direction of 0 °, 45 °, 90 ° or 135 °, the second step forms a symmetric matrix by adding the initial matrix with the transpose value, the third stage normalizes the matrix by dividing each matrix element by the number of pixel pairs then the resulting value will be used to produce 4 features of the gray level co-occurrence matrix (GLCM) ) namely energy, contrast, correlation, and entropy (Rizal, Gulo, Sihombing, Napitupulu, Gultom, & Siagian, 2019)
II. LITERATURE REVIEW
GLCM feature extraction is widely proposed by
researchers because it can be applied in various
problems such as Recognition of facial expressions
proposed by (Rizal, Gulo, Sihombing, Napitupulu,
Gultom, & Siagian, 2019) GLCM recognition rates
reach an average of 80%, (Sukiman, Suwilo, & Zarlis,
2019) proposed feature extraction using GLCM and
LVQ in facial recognition with a 90% recognition rate
and (Öztürka & Akdemir, 2018) applied the Feature
Extraction and Classification method for
Histopathological Images using GLCM, LBP,
LBGLCM, GLRLM, and SFTA .
2.1. Tomato Sauce
Tomato sauce is made from a mixture of tomatoes
and spices and then paste used is pink according to
the color of the tomato used, although it contains
large amounts of water the sauce has a long shelf life
because it contains acids, sugar, salt, and often given
preservatives (Ray, Saha, Raychaudhuri, &
Chakraborty, 2016)
2.2. Borax
Borax is a white crystalline powder containing boron,
borax is generally used for anti-fungi, wood
preservatives, and antiseptic ingredients.
Consumption of foods containing borax can cause
poisoning and can even cause death (Lathifah,
Turista, Azizah, & Khulaifi, 2019)
2.3. GLCM (Gray Level Co-Occurrence Matrix)
GLCM is a method that is often used in object
recognition because the method of glcm has quite
good steps in the process of object recognition, as for
the stages, namely: the initial stages of forming a
matrix of two-pixel pairs that align in the direction of
0 °, 45 °, 90 ° or 135 °, the second step forms a
symmetrical matrix by adding the initial matrix with
the transpose value, the third step normalizes the
matrix by dividing each matrix element by the
number of pixel pairs and then the resulting value will
be used to produce 4 features of the gray level co-
occurrence matrix (GLCM), namely energy, contrast,
correlation, and entropy (Rizal, Gulo, Sihombing,
Napitupulu, Gultom, & Siagian, 2019).
III. PROPOSED METHOD
3.1. Datasheet
Datasheet used in this study is the original tomato sauce image data and the image of tomato sauce that has been mixed with borax with different percentages starting from the image of the sauces with 70% borax content, the image of sauces with 50% borax content, the image of tomato sauce with borax content 30% and the image of tomato sauce that does not contain borax. A sampling of images using a camera phone brand Xiaomi Note 5 by mixing borax in the original sauce before the sample is used for the training and testing process.
3.2. Research Steps General research steps developed in this study can be seen in Figure 1.
Figure 1. General Research Steps for Tomato Sauce Recognition
Training
Original Image Read Pixel GLCM
Save Pattern
Model
Testing
Test Image
Original Image GLCM
Matching
Pattern Model
Output
Classification
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10508
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License. 8
In figure 1. there are two processes, namely the training process and the testing process, the input image training process is carried out pre-processing with grayscale and continued by extracting features from the tomato sauce image using GLCM and the extracted pattern model will be saved while in the testing phase, input image is preprocessing with grayscale and extracted using GLCM then proceeding to fit the pattern model matching, if the pattern is similar or close to the training pattern, the output of the classification is the result of the classification of tomato sauce image containing borax and not. The overall research steps to the introduction of saous tomato images containing borax and those not constructed in this study are illustrated in figure 2.
Figure 2. Overall Research Steps Introduction to Tomato Sauce Image
In Figure 2. can be seen after the system receives the input of the original tomato sauce image, the system will process the input image into grayscale, then the grayscale image is processed again using the steps in the GLCM method to generate weight values from the tomato sauce image and stored as a reference pattern for image classification. tomato sauce. The weight of the testing feature extraction will be matched to the weight of the training feature extraction using GLCM for recognition and is classified into the image of the tomato sauce containing borax or not.
IV. RESULT AND DISCUSSION
4.1. Tomato Sauce Image Samples
Samples of tomato sauce images containing borax and not used in this study amounted to 400 tomato sauce images with different percentages starting from the image of sauces with 70% borax content, image sauces with 50% borax content, image of tomato sauce with 30% borax content and image of tomato sauce that does not contain borax. A sampling of images using a camera phone brand Xiaomi Note 5 by mixing borax in the original sauce before the sample is used for the training and testing process. Figure 3. shows some sample images of tomato sauce used.
Figure 3 Examples of Several Samples of Tomato Sauce Image Used
Table 1. Test results for the classification of borax content 70% training and 30% testing
Original Borax
30%
Borax
50%
Borax
70%
Original 0.89 0 0.05 0.06
Borax
30%
0.02 0.94 0 0.04
Borax
50%
0 0.13 0.83 0.05
Borax
70%
0.05 0.09 0 0.86
In table 1, illustrate the results of the classification of borax content using GLCM with 70% training and 30% testing. In the original tomato sauce image GLCM is able to classify with an average accuracy rate of 89%, in the 30% content content borax GLCM is able to classify with an average accuracy rate of 94%, image content of 50% borax with an average accuracy rate of 83% and image 70% borax content with an average accuracy rate of 86%. The overall results of the classification system of borax content in tomato sauce using GLCM are implemented in Figure 4.
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10508
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License. 9
Figure 4. Overall Graph of Borax Content Classification Systemn On Tomato Sauce Using
GLCM
V. CONCLUSION AND SUGGESTION
The results showed that the gray level co-
occurrence matrix (GLCM) method was quite
optimal in classifying tomato sauce data containing
borax and not with an average percentage of the
introduction of 88%, to improve the quality of the
gray level co-occurrence matrix (GLCM)
performance. can add another approach to the
classification process. The addition of this approach
will certainly affect the computational speed and
accuracy of the classification process.
VI. REFERENCES
Indriani, O. R., Kusuma, E. J., Sari, C. A.,
Rachmawanto, E. H., & Setiadi, D. R.
(2017). Tomatoes Classification Using K-
NN Based on. International Conference on
Innovative and Creative Information
Technology (ICITech). Lathifah, Q. A., Turista, D. D., Azizah, L., &
Khulaifi, A. E. (2019). Identification of
formalin and borax on tuna in Ngemplak
market Tulungagung regency. Medical
Laboratory Analysis and Sciences Journal,
1(1).
Nkhata, S. G., & Ayua, E. O. (2018). Quality
attributes of homemade tomato sauce stored
at. African Journal of Food Science, 12(5).
Nursari, N., Karimuna, L., & Tamrin, T. (2016).
PENGARUH pH DAN SUHU
PASTEURISASI TERHADAP
KARAKTERISTIK KIMIA,
ORGANOLEPTIK DAN DAYA SIMPAN
SAMBAL. J. Sains dan Teknologi Pangan
(JSTP), 1(2), 151-158.
Öztürka, Ş., & Akdemir, B. (2018). Application of
Feature Extraction and Classification
Methods for Histopathological Image using
GLCM, LBP, LBGLCM, GLRLM and
SFTA. Procedia Computer Science.
Ray, S., Saha, R., Raychaudhuri, U., & Chakraborty,
R. (2016). DIFFERENT QUALITY
CHARACTERISTICS OF TOMATO
(SOLANUM LYCOPERSICUM) AS A
FORTIFYING INGREDIENT IN FOOD
PRODUCTS: A REVIEW. Technical
Sciences, 19(3).
Rizal, R. A., Gulo, S., Sihombing, O. D., Napitupulu,
A. B., Gultom, A. Y., & Siagian, T. J.
(2019). ANALISIS GRAY LEVEL CO-
OCCURRENCE MATRIX (GLCM)
DALAM MENGENALI CITRA
EKSPRESI WAJAH. Jurnal Mantik, 3(2).
Sukiman, T. S., Suwilo, S., & Zarlis, M. (2019).
Feature Extraction Method GLCM and LVQ
in Digital Image-Based Face Recognition.
SINKRON, 4(1).
Usman, N. B., Herawati, N., & Fitriani, S. (2019).
Quality of Sauce with Basic Ingredients of
Tomatoes, Carrots and Red Palm Oil.
JURNAL TEKNOLOGI PANGAN, 13(2).
Xie, C., Shao, Y., Li, X., & He, Y. (2015). Detection
of early blight and late. Scientific Reports.
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI https://doi.org/10.33395/sinkron.v4i2.10481
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative
Commons Attribution-NonCommercial 4.0 International License. 10
The Utilization of Expert System for
Diagnosing Diseases Cocoa Plants Based on
Android Using the Forward Chaining Method
Omar Pahlevi
Muhamad Kusumo Atmojo
Universitas Bina Sarana Informatika STMIK Nusa Mandiri
[email protected] [email protected]
Submitted: Feb 5, 2020
Accepted: Mar 9, 2020
Published: Apr 1, 2020
Abstract— Cacao plants originated from South America, then spread to North
America, Africa and Asia. In Indonesia, cocoa has been known since 1560, but has
become an important commodity since 1951. Cacao commodity plays an important role
in the national economy and is a national mainstay commodity. This shows that cocoa
is one of the results of plantation commodities that have a high economic value and play
an important role as a source of foreign exchange through exports, as well as
encouraging the regional economy, especially in rural areas. But behind the high value
of cocoa production, there are problems faced, including the low quality of cocoa in
Indonesia because cocoa plantations in Indonesia are threatened by pests and plant
diseases. Lack of information that is known by the plantation and cocoa farmers about
the types of diseases that attack cocoa plants, causing many cocoa plants that are not
handled properly. If this is allowed to continue it will impact on the declining quality
and production of cocoa plants. Current advances in information technology, especially
cellular phones, can be used as a means to improve public services, one of the results of
the development of cellular technology is the birth of cellular phones with the android
operating system. In this research produced if the symptoms data entered could not find
the type of cocoa plant disease because the input data did not match any disease data in
the database, the system would display the word "Can not find the disease you are
looking for because it is not related to fruit rot disease, stem cancer, vascular
antraknosem, streak dieback, upas fungus and root fungus ". From the data of symptoms,
diseases and relations above, the algorithm is depicted using a decision tree. Decision
tree is a picture of tracking symptoms, determining the disease and concluding results in
the form of a solution. In this application, using the Forward Chaining method so that
tracking begins with the selection of symptoms experienced then the results of the
diagnosis in the form of cocoa plant diseases.
Keywords— Expert System, Diagnosis, Cocoa Plant Disease, Android, Forward
Chaining Method
I. INTRODUCTION
At this time there is a development rapid in the field of science and technology, especially computer and communication technology or often referred to as the information age and communication technology (ICT). If in the beginning the computer was only used as a calculator, now the computer has been able to replace the role or complex tasks performed by humans, even being able to imitate human biological
processes in decision making called artificial intelligence (Listiyono, 2008).
Cocoa is the third most important commodity after rubber and palm oil, cocoa is one of the main sources of income for farmers in 31 provinces with the involvement of farmers totaling 1,539,401 heads of household. Increasing cocoa production in Indonesia is very rapid. In 1967 new production was 1,233 tons, in 2003 it reached 698,816 tons and in
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI https://doi.org/10.33395/sinkron.v4i2.10481
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License. 11
2010 it reached 837,918 tons cultivated by smallholder plantations (94.19%) with cocoa plant area reaching 1,650,621 Ha (Direktorat Jenderal Perkebunan, 2011).
Quality improvements and yields to support increased production and development of cocoa in Indonesia can be undertaken by using superior cocoa plant technology, information on land suitability for cocoa, plant propagation technology, pest and disease control major technologies, cloning technology, product processing technology, and technology cocoa industry development (Rubiyo & Siswanto, 2012).
Identifying the plant disease is not easy task, it needs experience and knowledge of plants and their diseases. Moreover, it requires accuracy in describing the symptoms of plant disesases. A person can depend on a system that posses experience and knowledge (expert systems) to enable him/or her identifying any type of disease. The methods that expert system uses differ from one system to another because that depends on the user’s primary knowledge of the case. Decision making depends mainly on the way of receiving that knowledge (Abu-Naser, Kashkash, & Fayyad, 2008).
Current technological developments have shown a lot of progress and provide benefits in various aspects of life, the use of technology is very helpful in various agencies such as government and private institutions, education in fulfilling the need to complete work. One of the technological developments that can be done is an expert system. Expert systems designed by utilizing information technology can assist in presenting accurate information. In addition, this technology has touched almost all mobile technology circles can be easily obtained. This is what underlies the need for an application regarding the disease diagnosis system in addition to overcoming obstacles and limitations in the world of health above. So that it is expected that this application can improve the performance of health services, and can reduce the emergence of hazards caused by symptoms of the disease because it can be detected more quickly (Tambunan, Siringoringo, Aruan, Aisyah, & Sitanggang, 2019).
In this case an important role of an expert is very relied upon to diagnose and determine the type of disease and provide examples of ways to overcome them to get the best solution. If there is a new disease found, an expert must conduct research to obtain information from the new disease and immediately provide information to farmers or farmer groups about the disease and how to overcome it (Koten, 2014).
Expert systems can help the activities of experts as experienced assistants and have the required knowledge. In its preparation, the expert system combines inference rules with certain knowledge
bases provided by one or more experts in a particular field. The combination of these two things is stored in a computer, which is then used in the decision making process for solving certain problems (Arifin, Slamin, & Retnani, 2017).
Today's technology has experienced very high progress, especially on smartphones today. The latest breakthrough in smartphones is marked by the emergence of an operating system, namely android (Sugiharja, Pahlevi, & Widyastuti, 2019). This can be supported by advances in existing information technology, especially mobile phones, can be used as a means to improve public services, one of the results of the development of cellular technology is the birth of cellular phones with the Android operating system. Android has various advantages as software that uses a computer code base that can be distributed openly (open source) so that users can create new applications in it. Therefore Android has a large community for application developers who expand functionality with the Android system (Kurniawan, 2015).
Android is an operating system for smartphones developed by Google. Android is opensource, so many android developers develop applications such as games, multimedia, file explorer, GPS, etc. including one of them in the development of expert systems (Susanti & Suhendri, 2017).
This research is related to previous research conducted by (Ariandi, Kurnia, Heriyanto, & Marry, 2019) with the title Expert System For Disease Diagnosis In Cocoa Plant Using Android-Based Forward Chaining Method. This research discusses about various types of pests and diseases that attack cocoa plants can cause losses especially for cocoa farmers. Identification of pests and diseases of cocoa plants must be done quickly and accurately, because these pests and diseases can quickly spread and attack cocoa plants in all plantation areas. The expert system of identifying cocoa pests and diseases with the forward chaining method is developed based on android or mobile that can be freely accessed by cocoa farmers or other users and is expected to help cocoa farmers in identifying and providing solutions to cacao pests and diseases so as to minimize losses which will be caused.
Then research conducted by (David Liauw, 2014) with the title Application of Forward Chaining in Expert Systems to Diagnose Corn Pests and Diseases. In the study explained that the expert system in general is a system that seeks to adopt human knowledge to the computer, so that computers can solve problems as is usually done by experts. Expert systems can collect and store the knowledge of an expert or several experts in a computer. Expert system software can help the work of an expert and can be used by farmers, laypeople to meet
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI https://doi.org/10.33395/sinkron.v4i2.10481
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License. 12
information about pests and diseases and as additional information for farmers who are just starting to grow corn but still lack knowledge about corn plants in this study create an expert system that using forward chaining method inference engine.
Next in the research conducted by (Qisty, 2018) with the research title Expert System for Identifying Pests and Diseases of Android-Based Cocoa Plants Using Forward Chaining and Certainty Factor Methods. In this study an expert system was built using the Forward Chaining inference method and the Certainty Factor calculation method. This expert system is made for Android device users. This study consisted of 70 symptom data, 6 pest data, 9 disease data, and 24 rules. When consulting, the user can answer Yes or No from the questions given by the system. User answers are then processed according to rules and are calculated using the certainty factor method. Testing is done to see whether the system can run well as expected. Testing consists of 3 aspects, namely system compatibility, the ability of the system to identify pests and diseases, and the benefits of the system according to targeted users. Based on these 3 tests, the cocoa expert system that has been developed has compatibility with the Android version of KitKat and Lollipop.
The algorithmic method used in this study is the same as the research conducted by (Hawa, Abdullah, & Usman, 2015) with the research title Expert System for Disease Diagnosis in Cocoa Plants Using the Forward Chaining Method (Case Study of the Indragiri Hilir Plantation Office). In this study describes a system that can provide information on several types and characteristics of diseases that interfere with cocoa. So that it can make it easy for farmers / users to find out how to diagnose diseases in cocoa plants. Of course this is expected to indirectly facilitate the diagnosis of farmers or users. In developing this system using the Forward Chaining method. From the results of the study it can be concluded that with the Expert System for Diagnosing Diseases of the Cocoa Plant, the Farmers can find out the disease that there are cocoa plants and the Farmers get a way to diagnose diseases in the cocoa plant.
After reviewing these studies, researcher made a study of expert systems for diagnosing diseases in the Android-based cocoa plant by using the Forward Chaining method. This application contains the introduction of the application and explanation of diseases and pests in cocoa plants. Disease in cocoa plants, how the application works and the features that are in the application.
II. LITERATURE REVIEW
A. Expert System
According to Durkin (1994) in (Sumpala &
Rasyid, 2019) defines expert system is one part of
artificial intelligence which has experienced rapid
development. In general, expert systems are systems
that try to adopt human knowledge to computers, so
that computers can solve problems like experts.
Expert system is a computer program designed to
model the ability to solve problems carried out by an
expert.
B. Cocoa Plants
According to (Layli, 2015) concluded that Cocoa
(Theobroma cacao L.) is one of the mainstay
commodity estates that has an important role for the
national economy, especially as a provider of
employment, a source of income, and the country's
foreign exchange.
C. Android According to (Yosef, 2014) explained that
Android is a Linux-based operating system that is used for cellular (mobile) phones such as shortcuts (smartphones) and tablet computers (PDAs). D. Forward Chaining Method
According to (Akil, 2017) The forward-chaining
algorithm is one of the two main methods of
reasoning when using an inference engine and can be
logically described as a repetition application from
the ponens mode (a set of valid inference rules and
arguments). The opposite of forward-chaining is
backward-chaining.
Then according to (Prambudi, Widodo, &
Widodo, 2018) in (Rusdiansyah, Setiawan, & Badrul,
2019) an inference that connects a multiplication
problem with solution called chain. A chain is sought
or is bypassed or crossed from a problem to obtain the
solution referred to as Forward Chaining. Another
way of describing the forward chaining this is by
reasoning from facts leading to the conclusion that
there is from the facts. In the advanced rules of
reasoning are tested one by one in a specific order.
The sequence may be either a sequence of rules of
incorporation into the base rules or also other
sequence determined by the user. Each time a rule is
tested, the expert system will evaluate whether
conditions are true or false. If the condition is true,
then the rule is kept then the next rule is tested.
Conversely, if conditions are wrong, it is not stored
and the next rule is tested. This process would be
repeated until the entire base of the rules tested with
a variety of conditions. Advanced reasoning work
with problems that started with the recording of the
initial information and the final settlement to be
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI https://doi.org/10.33395/sinkron.v4i2.10481
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License. 13
achieved, then the whole process will be done
sequentially.
III. PROPOSED METHOD
3.1. Object Research
This research was conducted by collecting data
through the results of data that have been collected
through official data available on the page
https://www.bps.go.id/
3.2. Systems Analysis and Design
The design process in this research includes the
design of the main systems and algorithms that will
be used in software performance. The knowledge
representation process is based on the rules that have
been obtained. Then translated into a decision table.
The data obtained is used to form a decision tree. The
symptoms of cocoa plant disease in this study were
brownish black spots, cocoa fruit felt soft and wet
when touched by a finger, skin of the bark was
slightly curved, often there was a reddish liquid, red
inner layer, brown spots on leaf bones, on the stem
will appear black binntik in the form of fungus thorns,
young Leaves Will Show Symptoms of Spots, brown
Necrosis, young fruit becomes wilted, dry, and
wrinkled, essence of the second or third leaf from the
point of growing yellow with spots green patches,
deciduous leaves so that there are symptoms of
toothless twigs, the presence of thin threads of fungus
such as silk on twigs, similar to spider nests, there is
a pink fungus crust, many leaves that remain attached
to the twigs even though it is dry, leaves turn yellow
withered and fall off, bald branches. The process of
gathering facts starts when the user enters the
consultation page which will display consultation
questions regarding the details of the disease. In this
menu the user can choose the answer yes or no to the
consultation question. After the user chooses, the
consultation answer will be processed.
Figure 1 Application Flowchart
3.3. Implementation
The expert system for diagnosing cocoa
diseases is implemented in the form of an android-
based software. The system was developed using the
forward chaining method by collecting symptom data
from users to find disease conclusions.
The main view of the system consists of user
pages and administrator pages. The main display for
the user consists of the main diagnosis features pages
and some additional features. The main view for
administrators consists of pages to update knowledge
in the system.
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI https://doi.org/10.33395/sinkron.v4i2.10481
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License. 14
Figure 2 Home Page Display
Figure 3 Menu Display
Figure 4 Consultation Page Views
Figure 5 Display of Consultation Results
Figure 6 Display Consultation Results If Not Found
Figure 7 Display of View Menu
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI https://doi.org/10.33395/sinkron.v4i2.10481
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License. 15
Figure 8 Display Symptom List
Figure 9 Display Detailed Data Symptoms
Figure 10 Disease List Display
Figure 11 Disease Detailed Data Display
IV. RESULT AND DISCUSSION
In designing this application, researcher used the
Forward Chaining algorithm. Where Forward
Chaining is a search method or tracking technique
that starts with information that is merging rules to
produce a conclusion or goal. This forward tracking
is very good if it works with problems that start with
recording the initial information and want to achieve
a final solution, because the whole process will be
done sequentially going forward.
The following table is a disease in cocoa plants
and their symptoms and their relationships that are
used as a comparison in designing applications in
research:
TABLE 1 Cocoa Plant Disease
Disease Code Remarks
P001 Fruit Rot Disease
P002 Stem Cancer
P003 Antraknose Disease
P004 Vascular Streak
Dieback Disease
P005 Upas Fungus Disease
P006 Root Fungus Disease
TABLE 2 Symptoms of Cocoa Plant Disease
Disease Code Remarks
G001 Blackish Brown spots
G002
Cocoa fruit feels soft
and wet if touched by a
finger
G003 Curved Bark
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI https://doi.org/10.33395/sinkron.v4i2.10481
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License. 16
G004 Often there is a reddish
liquid
G005 Red Inner Layer
G006 Brown spots on the leaf
bone
G007
Black stems appear on
the stems in the form of
mushroom spines
G008
Young Leaves Will
Show Symptoms of
Brown-colored
Necrosis Spots
G009 Young fruit withers,
dries, and wrinkles
G010
The second or third leaf
extract from the spot
grows yellow with
green patches
G011
The leaves fall so that
the symptoms appear
toothless twigs
G012
The presence of thin
threads of silk like
fungus on a branch,
similar to a cobweb
G013 There is a pink
mushroom crust
G014
Many leaves remain
attached to the branches
even though they are
dry
G015 The leaves turn wilted
and fall
G016 Bare branches
TABLE 3 Relationship Rules
Relationship IF (Disease Code) THEN
(Code)
1 G001,G002 P001
2 G003,G004,G005 P002
3 G006,G007,G008,G009 P003
4 G010,G011 P004
5 G012,G013,G014 P005
6 G015,G016 P006
If the entered symptom data cannot find the type
of cocoa plant disease because the data input does not match any disease data in the database, the system will display the word "Cannot find the disease you are looking for because it is not related to fruit rot disease, stem cancer, anthraxemic vascular, streak dieback, upas fungus and root fungus. From the data of symptoms, diseases and relations above, the
algorithm is depicted using a decision tree. The decision tree is a picture tracking the symptoms, determining the disease and concluding the results in the form of a solution. In this application, using the Forward Chaining method so that tracking begins with the selection of symptoms experienced then the results of the diagnosis in the form of types. Here is a decision tree diagram on the application in this research:
Figure 12 Decision Tree Diagram
This research uses cyclomatic white box testing.
Figure 13 Cyclomatic White Box Testing
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI https://doi.org/10.33395/sinkron.v4i2.10481
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License. 17
To provide a quantitative measurement of the
logical complex of the program the writer uses
cyclomatic complexity to guarantee that the number
of trials done is at least once done. The cyclomatic
complexity of the white box flow diagram can be
obtained by calculation : V(G) = E – N + 2.
Where :
E = The number of edges specified by the arrow
image
N = The number of node flowcharts is determined by
drawing a circle
So the results obtained are V(G) = 33 – 28 + 2 = 7.
Based on the above calculation, the number of
independent lines is 7. Below this is an independent
path that results from cyclomatic complexity:
1. 1-2-3-4-8-9-10-11-12-2
2. 1-2-3-4-5-13-14-17-18-19-20-17
3. 1-2-3-4-5-13-14-15-21-22-23-24-21
4. 1-2-3-4-5-13-14-15-16-2
5. 1-2-3-4-5-6-25-26-2
6. 1-2-3-4-5-6-7-2
7. 1-2-3-4-5-6-7-27-28
V. CONCLUSION AND SUGGESTION
5.1. Conclusion
Based on the design and implementation of an
application diagnosing android-based cocoa plant
diseases using the forward chaining algorithm
method in the previous chapters, the conclusion can
be drawn:
1. This application can help provide information
about diseases that attack cocoa plants and
provide controls to deal with these diseases.
2. This application is designed using the forward
chaining algorithm method in addition to
diagnosing cocoa plants, it also aims to find out
the symptoms of diseases in cocoa plants.
3. With this application many cocoa plants can be
saved and improve the quality of cocoa plants.
4. Adding insight for cocoa farmers to the
development of science and technology.
5.2. Suggestion
For further development of this research,
suggestions can be given as follows:
1. Add more data on types of diseases in cocoa
plants.
2. Display more pictures in each symptom question.
3. Add clearer and more detailed information about
cocoa plants.
4. Applications can later be developed on platforms
other than Android, such as iOS and Windows
Phone.
VI. ACKNOWLEDGMENT
We would like to take this opportunity to thank all
contributors of this journal. A special thanks to Dr.
Mochamad Wahyudi, MM, M.Kom, M.Pd as Rektor
of University of Bina Sarana Informatika for his
invaluable encouragement, guidance and support.
REFERENCES
Abu-Naser, S. S., Kashkash, K. A., & Fayyad, M.
(2008). Developing an Expert System for Plant
Disease Diagnosis. Journal of Artificial
Intelligence, 1(2), 78–85.
https://doi.org/10.3923/jai.2008.78.85
Akil, I. (2017). Analisa Efektifitas Metode Forward
Chaining Dan Backward Chaining Pada Sistem
Pakar. Jurnal Pilar Nusa Mandiri, 13(1), 35–
42.
Ariandi, V., Kurnia, H., Heriyanto, & Marry, H.
(2019). Expert System For Disease Diagnosis
In Cocoa Plant Using Android-Based Forward
Chaining Method. Journal of Physics:
Conference Series, 1339(1), 1–7.
https://doi.org/10.1088/1742-
6596/1339/1/012009
Arifin, M., Slamin, S., & Retnani, W. E. Y. (2017).
Penerapan Metode Certainty Factor Untuk
Sistem Pakar Diagnosis Hama Dan Penyakit
Pada Tanaman Tembakau. Berkala Sainstek,
5(1), 21.
https://doi.org/10.19184/bst.v5i1.5370
David Liauw. (2014). Penerapan Forward Chaining
Dalam Sistem Pakar Diagnosa Hama dan
Penyakit Tanaman Jagung. Seminar Nasional
Informatika, 90–95.
Hawa, S., Abdullah, & Usman. (2015). Sistem Pakar
Diagnosa Penyakit Pada Tanaman Kakao
Menggunakan Metode Forward Chaining
(Studi Kasus Dinas Perkebunan Indragiri
Hilir). Sistemasi, 4(2), 1–8.
Koten, Y. P. (2014). Sistem Pakar Fuzzy untuk
Mendiagnosa Penyakit Pada Tanaman Kakao
berbasis SMS Gateway. Universitas Atma Jaya
Yogyakarta.
Kurniawan, R. (2015). Sistem Pakar Identifikasi
Hama Dan Penyakit Tanaman Apel
Menggunakan Algoritma Forward Chaining
Berbasis Android. Universitas Muhammadiyah
Malang.
Layli, F. (2015). Evaluasi Kesesuaian Lahan Untuk
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI https://doi.org/10.33395/sinkron.v4i2.10481
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License. 18
Tanaman Kakao Di Kecamatan Selopuro
Kabupaten Blitar. 1, 1–13.
https://doi.org/10.1017/CBO9781107415324.0
04
Listiyono, H. (2008). Merancang dan Membuat
Sistem Pakar. Jurnal Teknologi Informasi
DINAMIK, XIII(2), 115–124.
Perkebunan, D. J. (2011). Pedoman Penanganan
Pasca Panen Kakao.
Qisty, A. Q. (2018). Sistem Pakar Identifikasi Hama
Dan Penyakit Tanaman Kakao Berbasis
Android Dengan Menggunakan Metode
Forward Chaining Dan Certainty Factor (Vol.
10). Universitas Lampung.
Rubiyo, & Siswanto. (2012). Peningkatan Produksi
dan Pengembangan Kakao ( Theobroma cacao
L. ) Di Indonesia. Buletin RISTRI, 3(1), 33–48.
https://doi.org/10.21082/jtidp.v3n1.2012.p33-
48
Rusdiansyah, Setiawan, S., & Badrul, M. (2019).
Diabetes Mellitus Diagnosis Expert System
With Web-Based Forward Chaining. SinkrOn,
3(2),61.
https://doi.org/10.33395/sinkron.v3i2.10055
Sugiharja, D., Pahlevi, O., & Widyastuti, R. (2019).
The RPTRA Geographic Information System
Application in Central Jakarta City Using the
Dijkstra Algorithm Based on Android.
SinkrOn,3(2),56.
https://doi.org/10.33395/sinkron.v3i2.10043
Sumpala, A. T., & Rasyid, R. (2019). Expert system
for corn plant disease diagnosis with the
breadth-first search method. IOP Conference
Series: Earth and Environmental Science,
382(1), 1–6. https://doi.org/10.1088/1755-
1315/382/1/012001
Susanti, D., & Suhendri. (2017). Perancangan Sistem
Pakar Diagnosa Penyakit Tanaman Mangga
Dengan Algoritma Depth First Search Berbasis
Mobile. Sintak, 24–32.
Tambunan, A. M., Siringoringo, S. R., Aruan, R.,
Aisyah, P. I., & Sitanggang, D. (2019). An
Expert System For Diagnosing Plant Diseases
Using Certainty Factor And Backward
Chaining Based On Android. Journal of
Physics: Conference Series, 1230(1), 1–7.
https://doi.org/10.1088/1742-
6596/1230/1/012075
Yosef. (2014). Android Black Box. Jakarta: Jasakom.
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10232
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative
Commons Attribution-NonCommercial 4.0 International License. 19
Analysis of Multi-attribute Utility Theory for
College Ranking Decision Making
Adidtya Perdana
Universitas Harapan Medan
Medan, Indonesia
Arief Budiman
Universitas Harapan Medan
Medan, Indonesia
Submitted: Oct 2, 2020
Accepted: Mar 17, 2020
Published: Apr 1, 2020
Abstract— Ranking of a tertiary institution, both state and private universities, can be
the basis of the tertiary institutions of interest to prospective new students. The better
the ranking of the college, the more popular the campus. In this study the author
discusses the case of campus ranking in the city of Medan where the results to be
received are the best campus decision making with the method used is the MAUT (Multi
Attribute Utility Theory) method. The aim is to see what results can be given by using
the MAUT method in determining the best campus in the city of Medan which results
in ranking the campus in Medan. Does it provide optimal results or not. But every case
that is solved using the methods in artificial intelligence, in this case the MAUT method
is a method of the Decision Support System, certainly provides optimal results even
though the results given are not complete or complete. Therefore, the writer has a vision
going forward, conducting research in this field, especially for the case of campus
ranking. In this study the variables used in determining campus ranking are Institutional,
Student Activities, Lecturer HR, Research and Community Service, and Innovation.
These five variables in the future can be added or subtracted as needed. The results
obtained are optimal ranking results but are still limited to the reference model for
internal institutions.
Keywords— Artificial Intelligence, Decision Support Systems, Multi Attribute Utility
Theory (MAUT), Campus Ranking.
I. INTRODUCTION
Ranking of a tertiary institution, both state tertiary institution (PTN) and private tertiary institution (PTS) can be a reference for these tertiary institutions to attract prospective students. The higher the rank of a tertiary institution, the more popular the campus is for prospective students to enter. To determine the ranking of a campus or university a cluster mapping scheme is needed under the auspices of the Ministry of Research, Technology and Higher Education to improve the quality of universities on a regular and sustainable basis. Quoted from the page ristekdikti.go.id there are 5 assessment components that are the basis for ranking universities under the auspices of the Ministry of Research, Technology and Higher Education including Human Resources Quality, Institutional Quality, Student Activity
Quality, Research and Community Service Quality, and Innovation Quality (Kemenristekdikti, 2018).
In the HR quality assessment includes the percentage of the number of lecturers based on the level of graduates, S1, S2 or S3, the percentage of the number of lecturers based on their rank and functional position, and the ratio of the number of students to lecturers. For the Institutional Quality assessment includes the accreditation of institutions and study programs, the number of internationally accredited study programs, the number of foreign students and the number of university collaborations. In the assessment of Student Activities only cover student performance (Kemenristekdikti, 2018).
At the fourth point, the evaluation of Research and Community Service Activities includes assessment of research performance, community service performance, and the number of scientific
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10232
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative
Commons Attribution-NonCommercial 4.0 International License. 20
articles (journals and proceedings) published both locally, nationally, and internationally and not indexed or indexed (Scopus, Thomson Reuters, Copernicus, etc.) per number of lecturers. And the last assessment of Innovation Quality that includes innovation performance (Kemenristekdikti, 2018).
To be able to calculate in detail how the ranking of the campus is carried out, we need a method that can process well and can provide optimal results. For this reason, the writer uses the decision support system method which is a branch of artificial intelligence, namely the Multi Attribute Utility Theory method or abbreviated as MAUT. Where this method is a scheme whose final evaluation of an object is defined as the weight added by a value that is relevant to its dimension value. One of the strengths of this method is that it makes it possible to make direct comparisons of various sizes with the final ranking ranking of evaluations that reflect the choices of decision makers (Satria, Atina, Simbolon, & Windarto, 2018).
From the research conducted by (Puspitasari, Rumita, & Pratama, 2013) the MAUT model or method is used to compare business strategy priorities by looking at the aspects of infrastructure, time, cost, and business opinion in their research on the problem of choosing a business strategy center case study the earthenware industry Kasongan, Bantul, Yogyakarta. In this research the use of the MAUT method for decision making and the results provided by this method is the utility value of the most optimal strategies including management training, product development and entrepreneurs meeting.
In the research conducted by (Badrul, Rusdiansyah, & Budihartanti, 2019) discussed the measurement of nutrition of children under five based on age, weight and height. Where the method used is Simple Additive Weighting (SAW) able to determine the nutritional status of toddlers by adding a toddler's body mass index variable so as to produce the right and valid decision using 20 toddler samples categorized by age group. In other research conducted by (Fajirwan, Arhami, & Amalia, 2018) discusses ranking or ranking several criteria in deciding someone is entitled to receive assistance to renovate a house based on survey data conducted directly in the field. The use of the MAUT method in this study is to process data that is entered in the survey directly in the field using a computerized system that produces the final results through ranking. From the results of the rating, twofa houses will be selected that will receive renovation assistance based on the highest value with the highest value limit with a value limit of ≥ 0.58. The boundary value ≥ 0.58 is obtained
from the results of discussions with the chairperson of Baitul Mal Aceh Barat Daya.
From the background explanation above, the writer raised the research theme of Analysis of Campus Ranking Results in Medan in Decision Making Using the Multi Attribute Utility Theory (MAUT) Method. It is expected that by using the MAUT method, in determining the ranking of campuses in the city of Medan will provide optimal results and in accordance with what is expected by the Kemenristekdikti cluster. And in the future the application made in the ranking of campuses in Medan using the MAUT method can be used by interested parties such as the Ministry of Research, Technology and Higher Education, as well as universities or campuses as a reference to improve the quality of each campus so that the rankings obtained can be ranked upgrading for the better.
II. LITERATURE REVIEW
2.1 Decision Support System
The concept of Decision Support System (DSS) was first revealed in 1971 by Michael Scoot Morton with the term Management Decision System. Then a number of companies, research institutions and universities began to conduct research and build a Decision Support System, so that the resulting production can be concluded that this system is a computer-based system aimed at assisting decision making in utilizing certain data and models to solve various problems that unstructured (Latif, Jamil, & Abbas, 2018).
Little defines Decision Support System as a computer-based information that produces a variety of alternative decisions to assist management in dealing with various structured and unstructured problems using data and models. From various definitions of a Decision Support System it can be concluded that a Decision Support System is a specific information system aimed at assisting management in making decisions relating to semi-structured issues where no one knows for certain how decisions should be made (Windarto, 2017) . This system has facilities to produce various alternatives that are interactively used by users (Latif et al., 2018). Decision support systems are composed by several components, namely the database, model base, and user interface (Imandasari & Windarto, 2017).
2.2 Multi Attribute Utility Theory Method
Multi Attribute Utility Theory (MAUT) is a final evaluation scheme, v (x) of an object x is defined as a weight added by a value that is relevant to its
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10232
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative
Commons Attribution-NonCommercial 4.0 International License. 21
dimension value. The phrase commonly used to refer to it is utility value. MAUT is used to convert several interests into numerical values on a scale from 0-1 with 0 representing the worst choice and 1 being the best. This allows direct comparisons of various sizes. The end result is a ranking order of evaluations that reflects the choices of decision makers. The whole evaluation value can be defined by the equation (Sadewo, Windarto, & Hartama, 2017) (Satria et al., 2018):
𝑉(𝑥) = ∑ 𝑊𝑗. 𝑋𝑖𝑗
𝑛
𝑖=1
Where V (x) is the evaluation value of an object to i and wi is the weight that determines the value of how important the element i to other elements. Whereas n is the number of elements. The total weighting is 1. In summary the steps in the MAUT method are as follows (Sadewo et al., 2017) (Satria et al., 2018):
1. Break a decision into different dimensions.
2. Determine alternative weights in each dimension.
3. List all alternatives
4. Enter the utility for each alternative according to its attributes.
5. Multiply the utility by the weight to determine the value of each alternative.
Matrix Normalization:
𝑈(𝑥) =𝑥 − 𝑥𝑖−
𝑥𝑖+ + 𝑥𝑖−
Information:
U(x) : Normalization Alternative Weight x
x : Alternative Weight
xi- : The worst weight (minimum) of the xth criterion
xi+ : The best weight (maximum) of the xth criterion
III. PROPOSED METHOD
3.1 Research Subject
The subjects in this study were campuses in the city of Medan to be ranked. The object of research is the ranking or ranking values of the campuses based on the criteria used. The variables used as rating criteria are:
1. HR Quality
2. Institutional Quality
3. Quality of Student Activities
4. Quality of Research and Community Service
5. Quality of Innovation
3.2 Data Used
In this study the data used to support the success of the study are as follows:
1. Campus or College Data in Medan City,
2. HR data such as the number of lecturers based on education and comparison of the percentage of the number of students.
3. Institutional Data such as Accreditation.
4. Student Activity Data owned by the college.
5. Research and Community Service Data based on cluster, and scientific publication data based on rank in Sinta2.
6. Innovation data owned by tertiary institutions is based on Ristekdikti innovation data.
For campus data used in this study using 10 campus data in the city of Medan. But to maintain the code of ethics, the names of the campuses are disguised using alphabetical order. So the campus data used are campus A, B, C, D, E, F, G, H, I, and J.
3.3 Analysis of Method Implementation
This section will explain how the Multi Attribute Utility Theory (MAUT) method is applied to this problem.
1. The first stage determines the weight of each criterion in which the criteria used are:
a. C1 = HR: Percentage of Lecturers and
Students
a. 1:15 – 1:20 : 4
b. 1:21 – 1:25 : 3
c. 1:26 – 1:35 : 2
d. < 15 or > 35 : 1
b. C2 = HR: Lecturer with Bechelor Degree
Education (Percentage)
a. 0% : 4
b. 0.1% - 0.99% : 3
c. 1% - 8% : 2
d. > 8% : 1
c. C3 = HR: Lecturer with Magister Degree
Education (Percentage)
a. 80% - 100% : 4
b. 60% - 79.99% : 3
c. 40% - 59.99% : 2
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10232
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative
Commons Attribution-NonCommercial 4.0 International License. 22
d. 0% - 39.99% : 1
d. C4 = HR: Lecturer with Doctoral Degree
Education (Percentage)
a. 10% - 15% : 4
b. 7% - 9.99% : 3
c. 3% - 6.99% : 2
d. 0% - 2.99% : 1
e. C5 = Institutional: Accreditation
a. A : 4
b. B : 3
c. C : 2
d. - : 1
f. C6 = Student Activities
a. > 3.00 : 4
b. 1.00 – 2.99 : 3
c. 0.1 – 0.99 : 2
d. 0 : 1
g. C7 = RCS: Research
a. Mandiri : 4
b. Utama : 3
c. Madya : 2
d. Binaan : 1
h. C8 = RCS: Community Services
a. Unggul : 4
b. Sangat Bagus : 3
c. Memuaskan : 2
d. Kurang Memuaskan : 1
i. C9 = RCS: Publication (Sinta Ratings)
a. 1 – 300 : 4
b. 301 – 500 : 3
c. 501 – 700 : 2
d. > 701 : 1
j. C10 = Inovation (Number of Inovation)
a. > 13 : 4
b. 8 – 12 : 3
c. 4 – 7 : 2
d. 0 – 3 : 1
2. The second stage determines the preference weights of each criterion, is:
a. C1 = HR: Percentage of Lecturers and Students = 3
b. C2 = HR: Lecturer with Bechelor Degree Education (Percentage) = 3
c. C3 = HR: Lecturer with Magister Degree Education (Percentage) = 3
d. C4 = HR: Lecturer with Doctoral Degree Education (Percentage) = 3
e. C5 = Institutional: Accreditation = 4
f. C6 = Student Activities = 1
g. C7 = RCS: Research = 3
h. C8 = RCS: Community Services = 3
i. C9 = RCS: Publication (Sinta Rating) = 3
j. C10 = Inovation (Number of Inovation) = 1
3.3 Manual Calculation
In this manual calculation the writer uses dummy data as an example of how this method works. The data used are 5 pieces that are represented using letters only (do not use the name of the actual college as an alternative name). The data are as follows:
TABLE I. ALTERNATIVE DATA CALCULATION
MANUAL
No Name
PT
C1 C2 C3 C4 C5 C6 C7 C8 C9 C10
1 A 32 2 70 11 B 0.33 Madya Memuaskan 109 2
2 B 30 5 102 14 B 0.92 Utama Sangat
Bagus
211 3
3 C 24 3 43 7 A 1.3 Mandiri Unggul 97 5
4 D 16 6 98 13 C 2.2 Binaan Memuaskan 102 3
5 E 21 2 89 10 B 1.3 Madya Memuaskan 189 2
Next calculate the weight of each data:
TABLE II. WEIGHTING RESULTS MATRIX
No Nama
PT
C1 C2 C3 C4 C5 C6 C7 C8 C9 C10
1 A 2 2 3 4 3 2 2 2 4 1
2 B 2 2 3 4 3 2 3 3 4 1
3 C 3 2 3 4 4 3 4 4 4 2
4 D 4 2 3 4 2 3 1 2 4 1
5 E 3 2 4 3 3 3 2 2 4 1
The following is a normalization matrix calculation:
1. Alternative Names : A (A1)
𝐴11 =2 − 2
4 − 2= 0
𝐴12 =2 − 2
2 − 2= 0
𝐴13 =3 − 3
4 − 3= 0
𝐴14 =4 − 3
4 − 3= 1
𝐴15 =3 − 2
4 − 2= 0.5
𝐴16 =2 − 2
3 − 2= 0
𝐴17 =2 − 1
4 − 1= 0.333
𝐴18 =2 − 2
4 − 2= 0
𝐴19 =4 − 4
4 − 4= 0
𝐴110 =1 − 1
2 − 1= 0
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10232
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative
Commons Attribution-NonCommercial 4.0 International License. 23
2. Alternative Names : B (A2)
𝐴21 =2 − 2
4 − 2= 0
𝐴22 =2 − 2
2 − 2= 0
𝐴23 =3 − 3
4 − 3= 0
𝐴24 =4 − 3
4 − 3= 1
𝐴25 =3 − 2
4 − 2= 0.5
𝐴26 =2 − 2
3 − 2= 0
𝐴27 =3 − 1
4 − 1= 0.667
𝐴28 =3 − 2
4 − 2= 0.5
𝐴29 =4 − 4
4 − 4= 0
𝐴210 =1 − 1
2 − 1= 0
3. Alternative Names : C (A3)
𝐴31 =3 − 2
4 − 2= 0.5
𝐴32 =2 − 2
2 − 2= 0
𝐴33 =3 − 3
4 − 3= 0
𝐴34 =4 − 3
4 − 3= 1
𝐴35 =4 − 2
4 − 2= 1
𝐴36 =3 − 2
3 − 2= 1
𝐴37 =4 − 1
4 − 1= 1
𝐴38 =4 − 2
4 − 2= 1
𝐴39 =4 − 4
4 − 4= 0
𝐴310 =2 − 1
2 − 1= 1
4. Alternative Names : D (A4)
𝐴41 =4 − 2
4 − 2= 1
𝐴42 =2 − 2
2 − 2= 0
𝐴43 =3 − 3
4 − 3= 0
𝐴44 =4 − 3
4 − 3= 1
𝐴45 =2 − 2
4 − 2= 0
𝐴46 =3 − 2
3 − 2= 1
𝐴47 =1 − 1
4 − 1= 0
𝐴48 =2 − 2
4 − 2= 0
𝐴49 =4 − 4
4 − 4= 0
𝐴410 =1 − 1
2 − 1= 0
5. Alternative Names: E (A5)
𝐴51 =3 − 2
4 − 2= 0.5
𝐴52 =2 − 2
2 − 2= 0
𝐴53 =4 − 3
4 − 3= 1
𝐴54 =3 − 3
4 − 3= 0
𝐴55 =3 − 2
4 − 2= 0.5
𝐴56 =3 − 2
3 − 2= 1
𝐴57 =2 − 1
4 − 1= 0.333
𝐴58 =2 − 2
4 − 2= 0
𝐴59 =4 − 4
4 − 4= 0
𝐴510 =1 − 1
2 − 1= 0
Matrix normalization results:
TABLE III. MATRIX NORMALITATION
N
o
Name
PT C1 C2 C3 C4 C5 C6 C7 C8 C9 C10
1 A 0 0 0 1 0.5 0 0.333 0 0 0
2 B 0 0 0 1 0.5 0 0.667 0.5 0 0
3 C 0.5 0 0 1 1 1 1 1 0 1
4 D 1 0 0 1 0 1 0 0 0 0
5 E 0.5 0 1 0 0.5 1 0.333 0 0 0
The next step will be to multiply between the normalization matrix and the preference weights. The following is the calculation of the matrix multiplication:
A1 = (3*0) + (3*0) + (3*0) + (3*1) + (4*0.5) + (1*0) + (3*0.333)
+ (3*0) + (3*0) + (1*0) = 5.999
A2 = (3*0) + (3*0) + (3*0) + (3*1) + (4*0.5) + (1*0) + (3*0.667)
+ (3*0.5) + (3*0) + (1*0) = 8.501
A3 = (3*0.5) + (3*0) + (3*0) + (3*1) + (4*1) + (1*1) + (3*0.1) +
(3*1) + (3*0) + (1*1) = 13.8
A4 = (3*1) + (3*0) + (3*0) + (3*1) + (4*0) + (1*1) + (3*0) + (3*0)
+ (3*0) + (1*0) = 7
A5 = (3*0.5) + (3*0) + (3*1) + (3*0) + (4*0.5) + (1*1) + (3*0.333)
+ (3*0) + (3*0) + (1*0) = 8.499
From the results of calculations performed using the MAUT method, campus ratings are obtained from the dummy data as follows:
TABLE IV. TABLE STYLES
Campus Score Rank
C 13.8 1
B 8.501 2
E 8.499 3
D 7 4
A 5.999 5
IV. RESULT AND DISCUSSION
In this section the author will describe the results obtained from the application of the MAUT method to the processed data. The data will later be used as criteria in the calculation process. Data obtained and processed are as follows:
TABLE V. DATA OBTAINED AND WILL BE
PROCESSED
N
o
Pergurua
n Tinggi
SDM
Persentase
Dosen
Mahasiswa
(1 : …)
Dosen
S1
Dose
n S2
Dosen
S3
1 A 50.8 1 201 32
2 B 42 18 450 70
3 C 15 8 293 41
4 D 17.7 14 166 7
5 E 33.4 9 148 13
6 F 64.6 4 86 14
7 G 45 10 289 32
8 H 37.7 46 265 50
9 I 34 2 223 25
10 J 1.1 3 57 1
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10232
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative
Commons Attribution-NonCommercial 4.0 International License. 24
N
o
Pergurua
n Tinggi
Kelembagaan
Akreditasi Perguruan Tinggi
1 A B
2 B A
3 C B
4 D C
5 E B
6 F C
7 G B
8 H B
9 I B
10 J -
N
o
Pergurua
n Tinggi
Kemahasiswaan
Nilai (Berdasarkan Nilai pada
pemeringkatan.ristekdikti.go.id)
1 A 0.074
2 B 0.708
3 C 0
4 D 0.226
5 E 0.163
6 F 0.034
7 G 0.129
8 H 0
9 I 0.094
10 J 0
N
o
Pergurua
n Tinggi
Penelitian, Pengabdian & Publikasi
Penelitian
Pengabdian
pada
Masyarakat
Publika
si (Sinta
Ranking
)
1 A Madya Memuaskan 288
2 B Utama Sangat Bagus 231
3 C Madya Memuaskan 110
4 D Madya Memuaskan 346
5 E Binaan Memuaskan 411
6 F Madya Memuaskan 362
7 G Madya Memuaskan 214
8 H Binaan
Kurang
Memuaskan 195
9 I Madya Memuaskan 242
10 J Binaan
Kurang
Memuaskan 1302
N
o
Pergurua
n Tinggi
Inovasi
Nilai (Berdasarkan
data.inovasi.ristekdikti.go.id)
1 A 0
2 B 0
3 C 0
4 D 0
5 E 0
6 F 0
7 G 0
8 H 0
9 I 0
10 J 0
Based on the data obtained, it can be seen the weighting results of all criteria from alternative data A1 to A10 (representation of campus data A, B, C, D, E, F, G, H, I and J) as follows:
TABLE VI. WEIGHTING ALL CRITERIA RESULTS
No Data
Alternatif C1 C2 C3 C4 C5 C6 C7 C8 C9 C10
1 A1 1 3 4 4 3 2 2 2 4 1
2 A2 1 2 4 4 4 2 3 3 4 1
3 A3 4 2 4 4 3 1 2 2 4 1
4 A4 4 2 4 2 2 2 2 2 3 1
5 A5 2 2 4 3 3 2 1 2 3 1
6 A6 1 2 4 4 2 2 2 2 3 1
7 A7 1 2 4 3 3 2 2 2 4 1
8 A8 1 1 3 4 3 1 1 1 4 1
9 A9 2 3 4 4 3 2 2 2 4 1
10 A10 1 2 4 1 1 1 1 1 1 1
After obtaining weighting matrix data from alternative data possessed the next step is to carry out the normalization process with the values 𝑥𝑖
− and 𝑥𝑖+
for each criterion as follows:
TABLE VII. WORST AND BEST WEIGHT OF THE X
CRITERIA
No Kriteria 𝒙𝒊− 𝒙𝒊
+
1 C1 1 4
2 C2 1 3
3 C3 3 4
4 C4 1 4
5 C5 1 5
6 C6 1 2
7 C7 1 3
8 C8 1 3
9 C9 1 4
10 C10 1 1
So we get the normalization matrix as follows:
TABLE VIII. NORMALIZATION MATRIX RESULTS
No Data
Alternatif C1 C2 C3 C4 C5 C6 C7 C8 C9 C10
1 A1 0 1 1 1 0.667 1 0.5 0.5 1 0
2 A2 0 0.5 1 1 1 1 1 1 1 0
3 A3 1 0.5 1 1 0.667 0 0.5 0.5 1 0
4 A4 1 0.5 1 0.333 0.333 1 0.5 0.5 0.667 0
5 A5 0.333 0.5 1 0.667 0.667 1 0 0.5 0.667 0
6 A6 0 0.5 1 1 0.333 1 0.5 0.5 0.667 0
7 A7 0 0.5 1 0.667 0.667 1 0.5 0.5 1 0
8 A8 0 0 0 1 0.667 0 0 0 1 0
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10232
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative
Commons Attribution-NonCommercial 4.0 International License. 25
9 A9 0.333 1 1 1 0.667 1 0.5 0.5 1 0
10 A10 0 0.5 1 0 0 0 0 0 0 0
The next step is to calculate the normalization matrix multiplication with the preference weights:
TABLE IX. NORMALIZATION MATRIX RESULTS
No Data Alternatif Result
1 A1 18.66667
2 A2 21.5
3 A3 19.16667
4 A4 15.83333
5 A5 14.66667
6 A6 14.83333
7 A7 16.16667
8 A8 8.666667
9 A9 19.66667
10 A10 4.5
From the results obtained from data processing and calculations using the Multi Attribute Utility Theory (MAUT) method, the results of the ranking of the best tertiary institutions in the city of Medan can be displayed:
TABLE X. RESULT RANKING USING MAUT
No Campus Score Ranking
1 B 21.5 1
2 I 19.66667 2
3 C 19.16667 3
4 A 18.66667 4
5 G 16.16667 5
6 D 15.83333 6
7 F 14.83333 7
8 E 14.66667 8
9 H 8.666667 9
10 J 4.5 10
From the results shown in table 4.17 it can be seen that University B was ranked 1 (first) with a value of 21.5 according to calculations using the MAUT method. Followed by University of I with a value of 19.6667 which was ranked 2, University C with a value of 19.1667 which was ranked 3, A with a value of 18.6667 which was ranked 4, and University G with a value of 161667 which was ranked 5.
Then University D with a value of 15.8333 which was ranked rank 6, University F with a value of
14.8333 which was ranked 7th, University E with a value of 14.6667 which was ranked 8th, University H with a value of 8.6667 which was ranked 9th, and University J with a value of 4.5 which was ranked 10th.
From the results provided by the calculation of the MAUT method there are certainly many shortcomings in terms of data that the author may not receive is too accurate and there are still many criteria that the authors and the team did not enter because of time and access limitations in collecting all the data that should have been used.
However, the application of this method can help, at least the institution where the writer is located, in seeing his ranking in the field and can be a benchmark of anything that can be changed or improved if you want to rank up and get the maximum value.
V. CONCLUSION AND SUGGESTION
The conclusions that can be drawn from the results of the research carried out are as follows:
1. With the implementation of the Multi Attribute Utility Theory method in the case of ranking the campus in the city of Medan able to provide optimal results based on predetermined criteria and weightings.
2. In this method the use of weighting criteria, weighting preferences, and alternative data is very influential in the calculation and results provided.
3. To be used as the main reference in ranking the campus in the city of Medan still cannot. But it can be used as an additional reference and can only be used by the internal campus of Universitas Harapan Medan due to the limited data available.
VI. ACKNOWLEDGMENT
Thank you to the research institute and journal publication at the Harapan University of Medan, Mr. Ruswan Nurmadi, who facilitated the authors in the research conducted so that the publication of this publication. And thanks also to the Chancellor of Universitas Harapan Medan, Prof. Prof. Dr. Ritha F. Dalimunthe and thanks Dr. Rahmat Widia Sembiring who helped in the completion of the research the author did. And thanks to all those who cannot be mentioned one by one.
VII. REFERENCES
Badrul, M., Rusdiansyah, R., & Budihartanti, C.
(2019). Application of Simple Additive
Weighting Method for Determination of
Toddler Nutrition Status. SinkOn, 4(1), 19–24.
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10232
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative
Commons Attribution-NonCommercial 4.0 International License. 26
Fajirwan, D., Arhami, M., & Amalia, I. (2018).
Sistem Pendukung Keputusan Penerimaan
Bantuan Renovasi Rumah Dhuafa
Menggunakan Metode Multi Attribute Utility
Theory. Jurnal Infomedia, 3(2), 49–57.
Imandasari, T., & Windarto, A. P. (2017). Sistem
Pendukung Keputusan dalam
Merekomendasikan Unit Terbaik di PDAM
Tirta Lihou Menggunakan Metode Promethee.
Jurnal Teknologi Dan Sistem Komputer, 5(4),
159.
https://doi.org/10.14710/jtsiskom.5.4.2017.159
-165
Kemenristekdikti. (2018). Kemenristekdikti
Umumkan Peringkat 100 Besar Perguruan
Tinggi Indonesia Non Vokasi Tahun 2018.
Retrieved February 12, 2019, from 17 Agustus
2018 website:
https://ristekdikti.go.id/kabar/kemenristekdikti
-umumkan-peringkat-100-besar-perguruan-
tinggi-indonesia-non-vokasi-tahun-2018/
Latif, L. A., Jamil, M., & Abbas, S. H. (2018).
SISTEM PENDUKUNG KEPUTUSAN TEORI
DAN IMPLEMENTASI. Sleman: Deepublish.
Puspitasari, N. B., Rumita, R., & Pratama, G. Y.
(2013). Pemilihan Strategi Bisnis dengan
menggunakan QSPM (Quantitive Strategic
Planning Matrix) dan Model MAUT (Multi
Atribute Utility Theory) (Studi Kasus pada
Sentra Gerabah Kasongan, Bantul,
Yogyakarta). Universitas Diponegoro, VIII(3),
171–180.
Sadewo, M. G., Windarto, A. P., & Hartama, D.
(2017). PENERAPAN DATAMINING PADA
POPULASI DAGING AYAM RAS
PEDAGING DI INDONESIA
BERDASARKAN PROVINSI
MENGGUNAKAN K-MEANS
CLUSTERING. InfoTekJar (Jurnal Nasional
Informatika Dan Teknologi Jaringan), 2(1),
60–67.
https://doi.org/10.30743/infotekjar.v2i1.164
Satria, E., Atina, N., Simbolon, M. E., & Windarto,
A. P. (2018). Spk : Algoritma Multi-Attribute
Utility Theory ( Maut ) Padadestinasi Tujuan
Wisata Lokal Di Kota Sidamanik. 3(2), 168–
172.
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10235
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative
Commons Attribution-NonCommercial 4.0 International License. 27
Decision Support System For Determining the
Best College High Private Using Topsis Method
Yuyun Dwi Lestari
Universitas Harapan Medan
Medan, Indonesia
Mardiana
Universitas Harapan Medan
Medan, Indonesia
Submitted: Oct 3, 2019
Accepted: Mar 15, 2020
Published: Apr 1, 2020
Abstract— In this study, the Best Private Universities in Medan will be discussed with
the TOPSIS method (Techniques for Other References with Similarities to Solutions).
TOPSIS is one method of Decision Support Systems that is used to accept decisions that
have many criteria. The general objective of this study is to use the TOPSIS method
which is expected to be able to provide optimal results in determining the best
universities. And specific targets of this study can provide estimates or indicators that
can be used by Universities in Medan in determining whether the College is the best or
not. The criteria used in this study were Institutional, Student Activities, Lecturer HR,
Research and Community Service, and Innovation. In the future these criteria can be
added according to the needs and requirements provided by the Ministry of Research,
Technology and Higher Education. By applying the TOPSIS Method to Decision
Supporters in the selection of the Best Private Universities in Medan City can provide
optimal results based on the criteria and weighting that has been determined.
Keywords— Decision Support System, TOPSIS, Higher Education, Ranking
I. INTRODUCTION
Ranking of a tertiary institution, both state and
private tertiary institutions, can be a reference for
these tertiary institutions to attract students. The
higher the rank of a tertiary institution, the more
popular the campus is for prospective students to
enter. To determine the ranking of a campus or
university a cluster mapping scheme is needed under
the auspices of the Ministry of Research, Technology
and Higher Education to improve the quality of
universities on a regular and sustainable basis.
Quoted from the page ristekdikti.go.id there are 5
assessment components that are the basis for ranking
universities under the auspices of the Ministry of
Research and Technology including Human
Resources Quality, Institutional Quality, Student
Activity Quality, Research and Community Service
Quality, and Innovation Quality (Kemenristekdikti,
2018) .
This research will create a Decision Support
System to determine the best private universities by
using the TOPSIS (Technique for Order of Preference
by Similarity to Ideal Solution) method. The number
of private universities in Medan by offering their
respective advantages will make prospective students
interested in getting into which private universities
are the best. Therefore, each tertiary institution needs
to adopt any indicators that are an assessment for
students so that they can be interested in entering the
tertiary institution.
TOPSIS has been used in many applications
including financial investment decisions, company
performance comparison, comparison in a specific
industry, operating system selection, customer
evaluation, and robot. (Muzakkir, 2017)
TOPSIS is a method for finding the ideal solution
based on the value of preference. The reason for using
the TOPSIS method is that in TOPSIS it uses the
concept of selected alternatives not only to have the
shortest distance from the positive ideal solution, but
also to have the longest distance from the negative
ideal solution. The concept of TOPSIS is simple and
easy to understand and has the ability to measure
alternative alternatives in mathematical form.
(Firdaus, Abdillah, & Renaldi, 2016).
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10235
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative
Commons Attribution-NonCommercial 4.0 International License. 28
Research conducted by (Erik Kurniawan,
Hindayati Mustafidah, 2015) with the title TOPSIS
Method for Determining Admission of New Medical
Education Students at the University of
Muhammadiyah Purwakerto, in the study that the
highest value of new prospective students can be
taken into consideration in the process of selecting
new student admissions at the Faculty of Medicine
Muhammadiyah University. After conducting
research into new student admissions, further
research can determine student achievement. This
research was conducted by (Herawatie & Wuryanto,
2017).
Other research conducted by (Saleh &
Information, 2016) this study determines the majors
for students conducted by the student section where
the assessment process refers to report cards, written
test results, and students' interests which are then
recapitulated and then calculated. Next (Santiary,
Ciptayani, Saptariani, & I Ketut Swardika, 2018) by
determining tourist sites in the city of Bali with the
TOPSIS method. There are also researchers who
conduct research using the AHP and TOPSIS
methods for smartphone brand selection. (Akmaludin
& Badrul, 2019)
II. LITERATURE REVIEW
2.1 Decision Support System
Decision support system is an interactive
computer-based information system, by processing
data with various models to solve unstructured
problems so that it can provide information that can
be used by decision makers in making a decision.
Decision making is a process of choosing an action
among several alternatives, so that the desired
objectives can be achieved. (Chamid, 2016)
The objectives of the decision support system are
as follows: (Badrul, Rusdiansyah & Budihartanti,
2019)
1. Assist in making decisions on structured problems
2. Providing support for the manager's consideration
and not intended to replace the manager's function
3. Increasing the effectiveness of decisions taken
more than improving efficiency.
4. The speed of computer computing enables
decision makers to do a lot of computing quickly
at a low cost
5. Increased productivity building a decision-
making group, especially experts, can be very
expensive
2.2 Technique For Order Preference By Similarity
To Ideal Solution (TOPSIS)
The TOPSIS method is widely used to complete
practical decision making. This is because the
TOPSIS method has a simple and easy to understand
concept, with efficient computing, and has the ability
to measure the relative performance of decision
alternatives in a simple mathematical form. In
general, the TOPSIS method process follows the
following steps (Kusumadewi, 2006)
1. Make a normalized decision matrix.
2. Make a normalized weighted decision matrix.
3. Determine a positive ideal solution matrix and
a negative ideal solution matrix.
4. Determine the distance between the values of
each alternative with a positive and negative ideal
solution matrix.
5. Determine the preference value for each
alternative.
TOPSIS method is also widely used to solve practical
decision problems and the concept is simple, easy to
understand, efficient computation and has the ability
to measure the relative performance of decision
alternatives in a simple mathematical form (Wahyuni
Industry, Khairunnisa, Abriyani, Muchlis, & Ulfa,
2017).
III. PROPOSED METHOD
3.1 Research Subject
The subjects in this study were campuses in the
city of Medan to be ranked. The object of research is
the ranking or ranking values of the campuses based
on the criteria used. The variables used as rating
criteria are:
1. HR Quality
2. Institutional Quality
3. Quality of Student Activities
4. Quality of Research and Community Service
5. Quality of Innovation
3.2 Data Used
In this study the data used to support the success
of the study are as follows:
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10235
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative
Commons Attribution-NonCommercial 4.0 International License. 29
1. Campus or College Data in Medan City,
2. HR data such as the number of lecturers based on
education and comparison of the percentage of the
number of students.
3. Institutional Data such as Accreditation.
4. Student Activity Data owned by the college.
5. Research and Community Service Data based on
cluster, and scientific publication data based on
rank in Sinta2.
6. Innovation data owned by tertiary institutions is
based on Ristekdikti innovation data.
3.3 Analysis of Method Implementation
This section will explain how the TOPSIS method
is applied in the selection of the best tertiary
institution in Medan. The first stage determines the
criteria and weight of each criterion in which the
criteria used are:
a. C1 = HR: Percentage of Lecturers and
Students
a. 1:15 – 1:20 : 4
b. 1:21 – 1:25 : 3
c. 1:26 – 1:35 : 2
d. < 15 or > 35 : 1
b. C2 = HR: Lecturer with Bechelor Degree
Education (Percentage)
a. 0% : 4
b. 0.1% - 0.99% : 3
c. 1% - 8% : 2
d. > 8% : 1
c. C3 = HR: Lecturer with Magister Degree
Education (Percentage)
a. 80% - 100% : 4
b. 60% - 79.99% : 3
c. 40% - 59.99% : 2
d. 0% - 39.99% : 1
d. C4 = HR: Lecturer with Doctoral Degree
Education (Percentage)
a. 10% - 15% : 4
b. 7% - 9.99% : 3
c. 3% - 6.99% : 2
d. 0% - 2.99% : 1
e. C5 = Institutional: Accreditation
a. A : 4
b. B : 3
c. C : 2
d. - : 1
f. C6 = Student Activities
a. > 3.00 : 4
b. 1.00 – 2.99 : 3
c. 0.1 – 0.99 : 2
d. 0 : 1
g. C7 = RCS: Research
a. Mandiri : 4
b. Utama : 3
c. Madya : 2
d. Binaan : 1
h. C8 = RCS: Community Services
a. Unggul : 4
b. Sangat Bagus : 3
c. Memuaskan : 2
d. Kurang Memuaskan : 1
i. C9 = RCS: Publication (Sinta Ratings)
a. 1 – 300 : 4
b. 301 – 500 : 3
c. 501 – 700 : 2
d. > 701 : 1
j. C10 = Inovation (Number of Inovation)
a. > 13 : 4
b. 8 – 12 : 3
c. 4 – 7 : 2
d. 0 – 3 : 1
1. The second stage determines the preference
weights of each criterion, is:
a. C1 = HR: Percentage of Lecturers and Students =
3
b. C2 = HR: Lecturer with Bechelor Degree
Education (Percentage) = 3
c. C3 = HR: Lecturer with Magister Degree
Education (Percentage) = 3
d. C4 = HR: Lecturer with Doctoral Degree
Education (Percentage) = 3
e. C5 = Institutional: Accreditation = 4
f. C6 = Student Activities = 1
g. C7 = RCS: Research = 3
h. C8 = RCS: Community Services = 3
i. C9 = RCS: Publication (Sinta Rating) = 3
j. C10 = Inovation (Number of Inovation) = 1
IV. RESULT AND DISCUSSION
At this stage will explain the results achieved
from the application of the TOPSIS method to the
data to be processed. These data will be used as
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10235
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative
Commons Attribution-NonCommercial 4.0 International License. 30
criteria in the calculation process. Data obtained and
processed are as follows:
TABLE I. DATA OBTAINED AND WILL BE
PROCESSED
No Perguruan
Tinggi
SDM
Persentase Dosen
Mahasiswa
(1 : …)
Dosen S1 Dosen
S2 Dosen S3
1 A 50.8 1 201 32
2 B 42 18 450 70
3 C 15 8 293 41
4 D 17.7 14 166 7
5 E 33.4 9 148 13
6 F 64.6 4 86 14
7 G 45 10 289 32
8 H 37.7 46 265 50
9 I 34 2 223 25
10 J 1.1 3 57 1
No Perguruan
Tinggi
Kelembagaan
Akreditasi Perguruan Tinggi
1 A B
2 B A
3 C B
4 D C
5 E B
6 F C
7 G B
8 H B
9 I B
10 J -
No Perguruan
Tinggi
Kemahasiswaan
Nilai (Berdasarkan Nilai pada
pemeringkatan.ristekdikti.go.id)
1 A 0.074
2 B 0.708
3 C 0
4 D 0.226
5 E 0.163
6 F 0.034
7 G 0.129
8 H 0
9 I 0.094
10 J 0
No Perguruan
Tinggi
Penelitian, Pengabdian & Publikasi
Penelitian Pengabdian pada
Masyarakat
Publikasi
(Sinta
Ranking)
1 A Madya Memuaskan 288
2 B Utama Sangat Bagus 231
3 C Madya Memuaskan 110
4 D Madya Memuaskan 346
5 E Binaan Memuaskan 411
6 F Madya Memuaskan 362
7 G Madya Memuaskan 214
8 H Binaan Kurang
Memuaskan 195
9 I Madya Memuaskan 242
10 J Binaan Kurang
Memuaskan 1302
No Perguruan
Tinggi
Inovasi
Nilai (Berdasarkan data.inovasi.ristekdikti.go.id)
1 A 0
2 B 0
3 C 0
4 D 0
5 E 0
6 F 0
7 G 0
8 H 0
9 I 0
10 J 0
From the data above we get the weighting matrix
results from the alternative data held as follows:
TABLE II. WEIGHTING ALL CRITERIA RESULTS
No Data
Alternatif C1 C2 C3 C4 C5 C6 C7 C8 C9 C10
1 A1 1 3 4 4 3 2 2 2 4 1
2 A2 1 2 4 4 4 2 3 3 4 1
3 A3 4 2 4 4 3 1 2 2 4 1
4 A4 4 2 4 2 2 2 2 2 3 1
5 A5 2 2 4 3 3 2 1 2 3 1
6 A6 1 2 4 4 2 2 2 2 3 1
7 A7 1 2 4 3 3 2 2 2 4 1
8 A8 1 1 3 4 3 1 1 1 4 1
9 A9 2 3 4 4 3 2 2 2 4 1
10 A10 1 2 4 1 1 1 1 1 1 1
After getting the weighting matrix results all the
next step criteria create a matrix xij consisting of m
alternatives and n criteria. This matrix contains the
weights / grade of each alternative to each of the
existing criteria:
TABLE III. CRITERIA DATA
No Kriteria Weight
1 C1 0,5
2 C2 0,5
3 C3 0,5
4 C4 0,5
5 C5 1
6 C6 0,25
7 C7 0,75
8 C8 0,75
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10235
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative
Commons Attribution-NonCommercial 4.0 International License. 31
9 C9 0,75
10 C10 0,25
Next, calculate the weight of each data used:
TABLE IV. NORMALIZATION MATRIX RESULTS
No Data
Alternatif C1 C2 C3 C4 C5 C6 C7 C8 C9 C10
1 A1 1 3 4 4 3 2 2 2 4 1
2 A2 1 2 4 4 4 2 3 3 4 1
3 A3 4 2 4 4 3 1 2 2 4 1
4 A4 4 2 4 2 2 2 2 2 3 1
5 A5 2 2 4 3 3 2 1 2 3 1
6 A6 1 2 4 4 2 2 2 2 3 1
7 A7 1 2 4 3 3 2 2 2 4 1
8 A8 1 1 3 4 3 1 1 1 4 1
9 A9 2 3 4 4 3 2 2 2 4 1
10 A10 1 2 4 1 1 1 1 1 1 1
Given a matrix x (1,1) the divisor is obtained by:
𝑟𝑖𝑗 =𝑥𝑖𝑗
√∑ 𝑥𝑖𝑗2𝑚
𝑖=1
After determining the divider value, the
normalized matrix is:
TABLE V. NORMALIZATION MATRIX RESULTS
No Alternatif C1 C2 C3 C4 C5 C6
1 A1 0.021739 0.06383 0.026144 0.033613 0.037975 0.064516
2 A2 0.021739 0.042553 0.026144 0.033613 0.050633 0.064516
3 A3 0.086957 0.042553 0.026144 0.033613 0.037975 0.032258
4 A4 0.086957 0.042553 0.026144 0.016807 0.025316 0.064516
5 A5 0.043478 0.042553 0.026144 0.02521 0.037975 0.064516
6 A6 0.021739 0.042553 0.026144 0.033613 0.025316 0.064516
7 A7 0.021739 0.042553 0.026144 0.02521 0.037975 0.064516
8 A8 0.021739 0.021277 0.019608 0.033613 0.037975 0.032258
9 A9 0.043478 0.06383 0.026144 0.033613 0.037975 0.064516
10 A10 0.021739 0.042553 0.026144 0.008403 0.012658 0.032258
C7 C8 C9 C10
0.055556 0.051282 0.032258 0.1
0.083333 0.076923 0.032258 0.1
0.055556 0.051282 0.032258 0.1
0.055556 0.051282 0.024194 0.1
0.027778 0.051282 0.024194 0.1
0.055556 0.051282 0.024194 0.1
0.055556 0.051282 0.032258 0.1
0.027778 0.025641 0.032258 0.1
0.055556 0.051282 0.032258 0.1
0.027778 0.025641 0.008065 0.1
After getting the normalization matrix results, the
results of the decision matrix table below can be
obtained from the results of the normalized decision
multiplied by the weight of Criteria Data with the
formula
𝑦𝑖𝑗 = 𝑤𝑖 . 𝑟𝑖𝑗
The results of the criteria weights are as follows
TABLE VI. NORMALIZATION WEIGTH
No Alternatif C1 C2 C3 C4 C5
1 A1 0.01087 0.031915 0.013072 0.016807 0.037975
2 A2 0.01087 0.021277 0.013072 0.016807 0.050633
3 A3 0.043478 0.021277 0.013072 0.016807 0.037975
4 A4 0.043478 0.021277 0.013072 0.008403 0.025316
5 A5 0.021739 0.021277 0.013072 0.012605 0.037975
6 A6 0.01087 0.021277 0.013072 0.016807 0.025316
7 A7 0.01087 0.021277 0.013072 0.012605 0.037975
8 A8 0.01087 0.010638 0.009804 0.016807 0.037975
9 A9 0.021739 0.031915 0.013072 0.016807 0.037975
10 A10 0.01087 0.021277 0.013072 0.004202 0.012658
C1 C2 C3 C4 C5
0.01087 0.031915 0.013072 0.016807 0.037975
0.01087 0.021277 0.013072 0.016807 0.050633
0.043478 0.021277 0.013072 0.016807 0.037975
0.043478 0.021277 0.013072 0.008403 0.025316
0.021739 0.021277 0.013072 0.012605 0.037975
0.01087 0.021277 0.013072 0.016807 0.025316
0.01087 0.021277 0.013072 0.012605 0.037975
0.01087 0.010638 0.009804 0.016807 0.037975
0.021739 0.031915 0.013072 0.016807 0.037975
0.01087 0.021277 0.013072 0.004202 0.012658
After the normalized decision matrix is obtained,
the results of the decision table below are the results
of the normalized weight decision divided by the
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10235
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative
Commons Attribution-NonCommercial 4.0 International License. 32
Maximal and Minimum normalized weights with the
formula:
𝐴+ = (𝑦1+, 𝑦2
+, … , 𝑦𝑛+)
𝐴− = (𝑦1−, 𝑦2
−, … , 𝑦𝑛−)
Min Max Results Based on criteria
TABLE VII. ALLTERNATIVE A+ DAN A-
A+ A-
0.045027 0.049612
0.0343 0.073038
0.033798 0.055858
0.040806 0.048788
0.053909 0.039176
0.051554 0.038186
0.046457 0.045095
0.07048 0.033601
0.037903 0.050788
0.07992 0.011129
Determine the distance between each alternative's
weighted value to the positive ideal solution and the
negative ideal solution. To determine the distance
between each alternative's weighted value to the
positive ideal solution, use the following equation
𝐷𝑖+ = √∑(𝑦𝑖
+
𝑛
𝑖=1
− 𝑦𝑖𝑗)2
Next to calculate the distance between the
weighted values of each alternative to the negative
ideal solution, the following equation is used
𝐷𝑖− = √∑(
𝑛
𝑖=1
𝑦𝑖𝑗 − 𝑦𝑖−)2
Alternative Results of D + and D-
TABLE VIII. ALLTERNATIVE D+ DAN D-
D+ D-
0.045027 0.049612
0.0343 0.073038
0.033798 0.055858
0.040806 0.048788
0.053909 0.039176
0.051554 0.038186
0.046457 0.045095
0.07048 0.033601
0.037903 0.050788
0.07992 0.011129
The final step is to calculate preferences for each
alternative, namely the results of the alternative D +
and D- added to the formula
𝑉𝑖 =𝐷𝑖
−
𝐷𝑖− + 𝐷𝑖
+
𝑉1 =0,049612
0,04502 + 0,049612= 0.52422
𝑉2 =0.04961
0.04503 + 0.04961= 0.68045
𝑉3 =0.07304
0.03430 + 0.07304= 0.62303
𝑉4 =0.05586
0.03380 + 0.05586= 0.54454
𝑉5 =0.04879
0.04081 + 0.04879= 0.42086
𝑉6 =0.03918
0.05391 + 0.03918= 0.42552
𝑉7 =0.03819
0.05155 + 0.03819= 0.49256
𝑉8 =0.04509
0.07048 + 0.04509= 0.32284
𝑉9 =0.05079
0.03790 + 0.05079= 0.57264
𝑉10 =0.01113
0.07992+0.01113= 0.12223
From the results of calculations carried out using the
TOPSIS method obtained ranks of the best tertiary
institutions in Medan City. The best results obtained
are the highest value, A2
V. CONCLUSION AND SUGGESTION
The conclusions obtained from the results of this
study are
1. With the application of the TOPSIS Method
in Decision Support in the selection of the
Best Private Universities in the City of
Medan is able to provide optimal results
based on predetermined criteria and
weighting.
2. In giving the criteria weights and preference
weights are very influential in the calculation
results, therefore in giving the value of
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10235
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative
Commons Attribution-NonCommercial 4.0 International License. 33
weights must be analyzed where the criteria
whose weight is preferred.
VI. ACKNOWLEDGMENT
Thank you to the research institute and journal
publication at the Harapan University of Medan, Mr.
Ruswan Nurmadi, who facilitated the authors in the
research conducted so that the publication of this
publication. And thanks also to the Chancellor of
Universitas Harapan Medan, Prof. Prof. Dr. Ritha F.
Dalimunthe and thanks Dr. Rahmat Widia Sembiring
who helped in the completion of the research the
author did. And thanks to all those who cannot be
mentioned one by one.
VII. REFERENCES
Akmaludin, A., & Badrul, M. (2019). Multi-criteria
for Selection of SmartPhone Brands Product
using AHP-TOPSIS Method. SinkrOn, 3(2),
154.
https://doi.org/10.33395/sinkron.v3i2.10069
Badrul, M., Rusdiansyah, R., & Budihartanti, C.
(2019). Application of Simple Additive
Weighting Method for Determination of
Toddler Nutrition Status. SinkrOn, 4(1), 19.
https://doi.org/10.33395/sinkron.v4i1.10145
Chamid, A. A. (2016). Prioritas Kondisi Rumah, 7(2),
537–544.
Erik Kurniawan, Hindayati Mustafidah, A. S. (2015).
Metode TOPSIS untuk Menentukan
Penerimaan Mahasiswa Baru Pendidikan
Dokter di Universitas Muhammadiyah
Purwokerto ( TOPSIS Method to Determine
New Students Admission at Medical School in
University of, III(November), 201–206.
Firdaus, I. H., Abdillah, G., & Renaldi, F. (2016).
Sistem Pendukung Keputusan Penentuan
Karyawan Terbaik. Seminar Nasional
Teknologi Informasi Dan Komunikasi 2016
(SENTIKA 2016), 2016(Sentika), 440–445.
https://doi.org/10.1111/gcb.13996
Gustri Wahyuni, E., Khairunnisa, N., Abriyani, F.,
Muchlis, N. F., & Ulfa, M. (2017). Sistem
Pendukung Keputusan Pemilihan Asisten
Terbaik Dengan Metode TOPSIS dan WP.
Teknoin, 22(2), 93–100.
https://doi.org/10.1016/j.marchem.2015.04.008
Herawatie, D., & Wuryanto, E. (2017). Sistem
Pendukung Keputusan Pemilihan Mahasiswa
Berprestasi dengan Metode Fuzzy TOPSIS.
Journal of Information System Engineering and
Business Interligence, 3(2).
Kadarsah Suryadi, A. ramdhani. (2015). Sistem
Pendukung Keputusan, (November), 1009–
1024.
Kemenristekdikti. (2018). Kemenristekdikti
Umumkan Peringkat 100 Besar Perguruan
Tinggi Indonesia Non Vokasi Tahun 2018.
Muzakkir, I. (2017). Penerapan metode topsis untuk
sistem pendukung keputusan penentuan
keluarga miskin pada desa panca karsa ii.
ILKOM Jurnal Ilmiah Volume 9 Nomor 3
Desember 2017, 9, 274–281.
Saleh, S., & Informasi, F. S. (2016). ( STUDI KASUS
YADIKA NATAR ), 16(02), 160–169.
Santiary, P. A. W., Ciptayani, P. I., Saptarini, N. G.
A. P. H., & I Ketut Swardika. (2018). Jurnal
Pengertian Topsis, 5(5), 621–628.
https://doi.org/10.25126/jtiik2018551120
Kusumadewi, S. H. (2006) ‘Fuzzy Multi-Attribute
Decision Making (Fuzzy MADM)’, Graha
Ilmu Yogyakarta.
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10502
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative
Commons Attribution-NonCommercial 4.0 International License. 34
Pears Classification Using Principal
Component Analysis and K-Nearest Neighbor
Moh. Arie Hasan
Sekolah Tinggi Mamajemen Informatika dan Komputer
Nusa Mandiri
Jakarta, Indonesia
Arief Setya Budi
Sekolah Tinggi Mamajemen Informatika dan
Komputer Nusa Mandiri
Jakarta, Indonesia
Submitted: Feb 20, 2020
Accepted: Mar 7, 2020
Published: Apr 1, 2020
Abstract— Pears is a fruit that is widely available in tropical climates such as in Western
Europe, Asia, Africa and one of them is Indonesia. There are many types of pears found
in Indonesia. Types of pears can be distinguished from the color, size, and shape. But it
is still difficult for ordinary people to get to know the types of pears. This is what gave
rise to the idea to make a study related to image processing to classify three types of
pears, namely abate, red and William pears in order to help determine the differences in
the three types of pears. The dataset used is 99 pears. The pear type classification process
is carried out by testing the pear image based on existing training data. Stages of training
and testing used consisted of image segmentation in the form of RGB and HSV
conversion for feature extraction. Furthermore, by using Principal Component Analysis
(PCA) data is grouped and K-Nearest Neighbor (KNN) is used to determine data
classification. The use of adequate training data will further improve the accuracy of the
classification of pears. The final results of this study indicate the accuracy of the
classification of pears for all three types of pears by 87.5%.
Keywords— Pears, Principal Component Analysis, Image Processing, K-Nearest Neighbor
I. INTRODUCTION
Pears are one of the many imported fruits found in the Indonesian market both in traditional markets and in supermarkets. This fruit has approximately 30 types, but in general there are only 3 types if distinguished by their skin color, namely green, yellow and red. In terms of shape, pears have a variety of shapes. Some pears are round like apples and some are shaped like bells (Octavia, Jesslyn, & Gasim, 2016). Because it has a shape that is almost similar to other fruits, many ordinary people find it difficult to classify a type of pear.
In image processing, computer graphics, and computer vision can be considered as "translating" input images into corresponding output images (Isola, Zhu, …, & 2017, n.d.). An important part of image processing is color. Besides being able to be seen visually, the image also has important information in the presentation of the image quality. External color
features and firmness of internal features are the most important factors observed by consumers (wholesalers or retailers) to determine fruit quality (Sehgal & Goel, 2016).
In a previous study, backpropagation neural networks were used with images incorporating grayscale, HSV, and L * a * b * to identify pear (Octavia et al., 2016). Research on Principal Component Analysis (PCA) has been carried out to identify floral patterns (Herfina, 2013) and application of distance transform method for identify handwriting pattern using PCA (Husein, 2019). Research using the K-Nearest Neighbor Method to perform Leaf Classification with Enhanced Image Features (Liantoni, 2016) and Fruits Recognition based on Texture Features and K-Nearest Neighbor (Ariffin, Mustaffa, Abdullah, & Nasharuddin, 2018).
Based on these problems, we need a way to classify three types of pears namely abate, red and
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10502
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative
Commons Attribution-NonCommercial 4.0 International License. 35
william pears using image processing. The process of classifying pears using PCA and KNN methods. The purpose of this study is to classify the types of abate, red and william pears from existing images. Furthermore, the image is processed using the matlab application to get the results of image classification of abate, red and william pears.
II. LITERATURE REVIEW
Principal Component Analysis (PCA)
Principal component analysis (PCA) is a well-
established method for feature extraction and
dimensionality reduction (Subasi & Gursoy, 2010).
PCA is a method that involves a mathematical
procedure that changes and transforms a large
number of correlated variables into a small number of
uncorrelated variables, without losing important
information in them. A number of two-dimensional
images of each three-dimensional object that will be
recognized, collected to represent the object as a
reference image. From the set of reference images,
feature extraction will then be performed to obtain
characteristic information (characteristics) of the
object. The result of feature extraction is used for
multi object orientation recognition process.
Principal Component Analysis is widely used to
project or convert a large data set into a form of data
presentation with a smaller size. PCA transformation
to a large data space will produce a number of
orthonormal basis vectors in the form of a collection
of eigenvectors from a particular covariant matrix
that can optimally present the data distribution
(Muhammad & Isnanto Riza, n.d.). The concept of
using PCA includes the calculation of standard
deviation values, covariance matrices, eigenvalue
values and eigen vectors. PCA can use the method of
warranty or correlation. If needed, the data is
standardized first so it approaches the standard
normal distribution. In this case the covariance
method is used with the following algorithm:
1. Collecting data in the form of gray-level matrix
X of size M x N. Suppose that is a vector
N x 1.:
2. Calculate average:
𝑥 = 1
𝑀∑ 𝑥𝑖
𝑀
𝑖=1
(1)
3. Calculate the average difference:
Φi = xi - x
(2)
4. Determine the covariance matrix From matrix
X=[Φ1 Φ2 … ΦM] (matriks NxM), Hitung
kovarian:
𝐶 = 1
𝑀∑ 𝜙𝑛𝜙𝑛
𝑇 = 𝑋𝑋𝑇
𝑀
𝑛=1
(3)
5. Determine the characteristic values and
characteristic vectors of the covariance matrices
C : 𝜆1 > 𝜆2 > ………. > 𝜆N (4)
and
C : u1, u2,……………., un (5)
6. Sort the characteristic vector u and the
characteristic value λ in the diagonal matrix in
descending order according to the greatest
cumulative probability value for each
characteristic vector so that the dominant
characteristic values are obtained (Herfina, 2013).
The use of the PCA method aims to group data
into several classes which are then grouped so that
they can classify images of abate, red and william
pears.
K-Nearest Neighbor (KNN)
K-Nearest Neighbor (KNN) algorithm is a
method for classifying objects based on learning data
that is the closest distance to the object. Learning data
is projected into multi-dimensional space, where each
dimension is representing features of data. The
purpose of the KNN algorithm is to classify new
objects based on attributes and training samples
where the results of the new test samples are
classified based on the majority of the categories in
the KNN. In the classification process, this algorithm
does not use any model to be matched and only based
on memory (Liantoni, 2016).
The working principle of the KNN is to find the
closest distance between the data to be evaluated with
its closest neighbor K in the training data. The
training data is projected into a multi-dimensional
space, where each dimension represents the features
of the data. This space is divided into sections based
on the classification of training data. A point on this
space is marked by class c, if class c is the most
common classification found in the nearest k of the
point. Near or far neighbors are usually calculated
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10502
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative
Commons Attribution-NonCommercial 4.0 International License. 36
based on Euclidean distances with the following
formula:
𝑑𝑖 = √∑(𝑥2𝑖 − 𝑥1𝑖)2
𝑝
𝑖=1
(6)
With x1 = sample data, x2 = test data, i = data
variable, dist = distance, p = data dimension. In the
learning phase, this algorithm only stores feature
vectors and classifications of learning data. In the
classification phase, the same features are calculated
for the test data. The distance of this new vector to all
the learning data vectors is calculated, and the closest
number of k is taken. The new points are predicted to
be included in the most classifications of these points.
The best k value for this algorithm depends on the
data. Generally, high k values reduce the effect of
noisation on classification, but make the boundaries
between each classification more blurred. A good k
value can be chosen with parameter optimization, for
example by using cross-validation. A special case
where classification is predicted based on the closest
learning data (in other words, k = 1) is called the
nearest neighbor algorithm (Whidhiasih, Wahanani,
& Supriyanto, 2013).
Evaluation of KNN Performance
Confusion matrix is used to evaluate the
performance of an algorithm. Confusion matrix has
information about the actual data and the results of
the prediction of a classification into matrix form
(Pulungan, Zarlis, & Suwilo, 2019).
III. PROPOSED METHOD
The data used in this study are 99 images with
75% of the dataset used for training, and 25% for
testing. Dataset consisted of 75 images of pear images
consisting of 25 training data of abate pear images, 25
training data of red pear images, and 25 training data
of pear william images. The test data consisted of 8
images of pear abate, 8 images of pear red and 8
images of pear william (“Fruits 360 | Kaggle,” n.d.).
Examples of images of these three types of fruit can
be seen in Figure 1.
(a) (b) (c)
Figure 1. Abate Pear (a), Red Pear (b), and William
Pear (c)
TABLE I. IMAGE DATASET OF PEAR
Class Number
of Images
Training
Images
Testing
Images
Abate 33 25 8
Red 33 25 8
William 33 25 8 The classification process of pear image types can be
seen in Figure 2.
Figure 2. Classification Design of Pears
The picture shows that the method used for the classification of pear image types starts from the input of pear image then image segmentation is carried out to get the results of segmentation. Then the feature extraction process becomes red, green, blue, hue, saturation, and value. The feature extraction results that have been obtained are converted into principal components. The next step is to classify the KNN algorithm to determine the three types of pear images.
Input RGB Image
RGB color space is widely used and is usually the default color space for storing and representing digital images. We can get other color spaces from RGB or non-linear transformations. RGB color space is the color space used by computers, graphics cards and monitors or LCDs
Input Image
Segmentation
Feature Extraction
Classification K-Nearest Neighbor (KNN)
Data Distribution of Classification Results
Principal Component Analysis (PCA)
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10502
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative
Commons Attribution-NonCommercial 4.0 International License. 37
(Kolkur, Kalbande, Shimpi, …, & 2017, n.d.). This process aims to display the RGB (Red, Green, Blue) color space of the managed pear image.
RGB formula:
𝑟 =𝑅
𝑅+𝐺+𝐵 (7)
𝑔 =𝐺
𝑅+𝐺+𝐵 (8)
𝑏 =𝐵
𝑅+𝐺+𝐵 (9)
Image Segmentation Image segmentation is part of the image processing
process to divide an image into homogeneous regions based on certain similarity criteria between the gray level of a pixel and the gray level of its neighboring pixels, then the results of this segmentation process will be used for further processing. The Otsu method is a method for segmenting digital images using threshold values automatically, i.e. changing gray digital images to black and white based on comparison of threshold values with pixel color values of digital images. To get the threshold value there is a calculation that must be done. The first step that must be done is to make a histogram. From the histogram we can know the number of pixels for each gray level. The gray level of the image is expressed as i through L. The level i starts with 1, which is pixel 0. For L, the maximum level is 256 with pixels worth 255 (Syafi’i, Wahyuningrum, & Muntasa, 2016).
The threshold value to look for in a grayscale image is expressed as k. The value of k ranges from 0 to L-1, with a value of L = 256. So the probability of each pixel at level i is expressed by the equation :
𝑃𝑖 = 𝑛𝑖
𝑁 (10)
The cumulative number formula of (k) , for L = 0, 1, 2, ..., L-1:
𝜔(𝑘) = ∑ 𝑝𝑖
𝑘
𝑖=0
(11)
The cumulative average formula of (k) , for L = 0, 1, 2, ..., L-1:
𝜇(𝑘) = ∑ 𝑖. 𝑝𝑖
𝑘
𝑖=0
(12)
Formula for calculating the mean global intensity k T :
𝜇𝑇(𝑘) = ∑ 𝑖. 𝑝𝑖
𝐿−1
𝑖=0
(13)
The equation for between class variance :
𝜎𝐵2(𝑘) =
[𝜇𝑇𝜔(𝑘) − 𝜇(𝑘)]2
[𝜔(𝑘)[1 − 𝜔(𝑘)]
(14)
The results of the calculation between the variance class look for the maximum value. The largest value is used as the threshold or the value of (k), with the equation
𝜎𝐵2(𝑘∗) = 𝑚𝑎𝑥1≤𝑥≤𝐿𝜎𝐵
2(𝑘) (15)
Between class variance aims to find the threshold value of a grayscale image, the threshold value is used as a reference value to convert a grayscale image to a binary image. Each image has a different threshold value [6].
Hue Saturation Value (HSV)
Input images in the RGB color space are converted to HSV color space using transformations. HSV images are collections of three different images as hue, saturation, and value (Shaik, Ganesan, Kalist, …, & 2015, n.d.). HSV has a closeness to the RGB system in describing colors that humans can see. HSV serves to reduce the intensity of light from outside and be able to detect certain objects. Here's the formula from RGB to HSV:
Cmax = max(R’ G’ B’) Cmin = min(R’ G’ B’)
∆ = Cmax - Cmin
(16)
Cmax functions to determine the largest constant value
in the RGB value, while Cmin determines the smallest
value in the RGB value.
Plotting Data Distribution
Plotting the data distribution is done with the aim of
testing and viewing the image data distribution graph which
is processed based on hue and saturation values to see the results of testing accuracy of pear type image processing.
Plotting the data distribution that will be displayed is the
distribution of training data in each class, the distribution of training data for each class along with the boundary lines
and and the distribution of test data in each class.
IV. RESULT AND DISCUSSION
Research tool specifications The tools used on this research is as follows :
1. Laptop with processor : AMD E-350 1.60GHz
2. Operating system : Windows 8 32-bit
3. Software Matlab R2013a used for making programs
classification of determining the type of pears
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10502
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative
Commons Attribution-NonCommercial 4.0 International License. 38
The initial process of pear type classification is input of
pear image from the training data in this study. Training
data can be seen in Figure 3.
Figure 3. Image of Pear Train Data
Furthermore, the data that has been inputted is
performed image segmentation using morphological
operation methods to perfect the results of segmentation. Convert grayscale image that aims to determine the
foreground area and background area with the value of a
binary image, as shown in figure 4.
Figure 4. Binary Image
Then the transformation of the color space from the
RGB image to the HSV image (Hue, Saturation, Value) is
used as a reference to recognize the color of an object in a digital image and reduce the intensity of light from outside
which can be seen in Figure 5.
Figure 5. Image of the Segmentation Results
Based on the image results of segmentation that has been obtained, then feature extraction is performed to
obtain the value of RGB, hue, saturation, value, and area of
the pear image. After that the reduction is done using the
PCA algorithm to get the results of the classification of images of abate, red and william pears.
The test results can be seen from plotting graphs of
processed image data. Plotting the distribution of training
data in each class is shown in Figure 6.
Figure 6. Distribution of Training Data
Based on the training data that has been obtained,
testing is done using test data. The following is a display of the distribution of training data and test data based on the
boundary line using the PCA and KNN algorithms seen in
Figure 7.
Figure 7. Distribution of Test Data and Training Data
Furthermore, to get the accuracy of the classification of
pears, testing is done using the GUI application using the
Matlab application. The designed application consists of
several functions, namely image input, image segmentation
process, feature extraction, and the process of determining
the results of classification. Following is the appearance of
the Matlab GUI Application that has been made:
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10502
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative
Commons Attribution-NonCommercial 4.0 International License. 39
Figure 8. Display of Image Input Matlab GUI Application
Following is the appearance of the Matlab GUI
Application segmented images, which consist of binary
images and segmented images:
Figure 9. Display of Matlab GUI Application Image
segmentation results
The following is the display of the extracted image
features consisting of Red, Green, Blue, Hue, Saturation, Value, and Area features in the Matlab GUI Application:
Figure 10. Display of Matlab GUI Application Extraction
Results
Based on the feature extraction results that have been
obtained, a process is carried out to determine the
classification using the PCA and KNN algorithms. The
following is the appearance of the Matlab GUI
classification results of the three types of pears:
Figure 11. Display of Matlab GUI Application Results of
Abate Pear Classification Results
Figure 12. Display of Matlab GUI Application Results of
Classification of Red Pears
Figure 13. Display of Matlab GUI Application Results of
William Pear Classification Results
The test results carried out on a test data of 24 pears
using the Matlab GUI application can be seen in table 1.
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10502
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative
Commons Attribution-NonCommercial 4.0 International License. 40
TABLE 2. PEARS IMAGE PROCESSING RESULTS
No. Test Image
Original Class
Output Class
Result
1
Abate Abate True
2
Abate Abate True
3
Abate Abate True
4
Abate Abate True
5
Abate Abate True
6
Abate Abate True
7
Abate Abate True
8
Abate Abate True
9
Red Red True
10
Red Red True
11
Red Red True
12
Red Red True
13
Red Red True
14
Red Red True
15
Red Red True
16
Red Red True
17
William William True
No. Test Image
Original Class
Output Class
Result
18
William Abate False
19
William Abate False
20
William Abate False
21
William William True
22
William William True
23
William William True
24
William William True
Based on the table above, 21 pears were successfully
classified according to their type, but there were 3 William Pears classified as Pear Abate. This can be seen in the GUI
display Figure 14
Figure 14. Display of the Matlab GUI Application Results of William Pear Classification Classified into Abate Pear
Image data used for testing are 24 consisting of 8
images of pear abate, 8 images of red pear, and 8 images of pear william. The test results show that the number of pears
that were successfully classified according to their class
amounted to 21 images while 3 images did not match.
Classification results will be presented in the form of confusion matrix. This table consists of predict classes and
actual classes. The 3x3 confusion matrix model is shown in
Table 3 while the accuracy value of the model is obtained
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10502
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative
Commons Attribution-NonCommercial 4.0 International License. 41
from equation (17), the exact amount of data classified
divided by the total data. Based on the calculation results
obtained an accuracy value of 87.5%. The resulting
accuracy indicates that the PCA and KNN algorithms are very well applied in the classification of fruit types.
TABLE 3. CONFUSION MATRIX
ACCURACY = 87.5% Actual
Class Predict Class Total
Abate
(A)
Red
(B)
William
(C)
Abate (A) 8 0 0 8
Red (B) 0 8 0 8
William (C) 3 0 5 8
11 8 5 24
Akurasi = 𝐴𝐴+𝐵𝐵+𝐶𝐶
𝐴𝐴+𝐴𝐵+𝐴𝐶+𝐵𝐴+𝐵𝐵+𝐵𝐶+𝐶𝐴+𝐶𝐵+𝐶𝐶
(17)
V. CONCLUSION AND SUGGESTION
This study found the results of how ordinary people can easily determine the type of pear only from an image processing. From the results of the classification process of image processing of abate, red and william pears, an accuracy of 87.5% was obtained. Using the Principal Component Analysis and K-Nearest Neighbor algorithm is very suitable in the classification process of pears. Image quality is very influential on the results of classification as well as the amount of training data used to obtain classification results. The more training data used, the better the accuracy of the classification of pears. It is recommended to develop further research using more than three types of pears.
VI. REFERENCES
Ariffin, N. I. K., Mustaffa, M. R., Abdullah, L. N., &
Nasharuddin, N. A. (2018). Fruits Recognition
based on Texture Features and K-Nearest
Neighbor. International Journal of
Engineering & Technology, 7(4.31), 452–458.
Fruits 360 | Kaggle. (n.d.). Retrieved August 22,
2019, from
https://www.kaggle.com/moltean/fruits
Herfina. (2013). Pengenalan Pola Bentuk Bunga
Menggunakan Principle Component Analysis.
Seminar Nasional Teknologi Informasi Dan
Multimedia, (7), 25–30.
Husein, A. M. (2019). Penerapan Metode Distance
Transform Pada Kernel Discriminant Analysis
Untuk Pengenalan Pola Tulisan Tangan Angka
Berbasis Principal Component Analysis .
SinkrOn, 2(SinkrOn), 31–36.
Isola, P., Zhu, J., … T. Z.-P. of the I., & 2017,
undefined. (n.d.). Image-to-image translation
with conditional adversarial networks.
Openaccess.Thecvf.Com.
Kolkur, S., Kalbande, D., Shimpi, P., … C. B.
preprint arXiv, & 2017, undefined. (n.d.).
Human skin detection using RGB, HSV and
YCbCr color models. Arxiv.Org.
Liantoni, F. (2016). Klasifikasi Daun Dengan
Perbaikan Fitur Citra Menggunakan Metode K-
Nearest Neighbor. Jurnal ULTIMATICS, 7(2),
98–104. https://doi.org/10.31937/ti.v7i2.356
Muhammad, J., & Isnanto Riza, S. I. (n.d.).
Identifikasi iris mata menggunakan metode
analisis komponen utama dan perhitungan jarak
euclidean, 1–9.
Octavia, M., Jesslyn, K., & Gasim. (2016).
Perbandingan Tingkat Akurasi Jenis Citra
Keabuan , HSV, Dan L*a*b* Pada Identifikasi
Jenis Buah Pir. Ilmiah Informatika Global,
7(1), 7–11.
Pulungan, A. F., Zarlis, M., & Suwilo, S. (2019).
Analysis of Braycurtis, Canberra and Euclidean
Distance in KNN Algorithm. SinkrOn, 4(1), 74.
https://doi.org/10.33395/sinkron.v4i1.10207
Sehgal, P., & Goel, N. (2016). Auto-annotation of
tomato images based on ripeness and firmness
classification for multimodal retrieval. In 2016
International Conference on Advances in
Computing, Communications and Informatics
(ICACCI) (pp. 1084–1091). IEEE.
https://doi.org/10.1109/ICACCI.2016.7732189
Shaik, K., Ganesan, P., Kalist, V., … B. S.-P. C., &
2015, undefined. (n.d.). Comparative study of
skin color detection and segmentation in HSV
and YCbCr color space. Elsevier.
Subasi, A., & Gursoy, M. I. (2010). EEG signal
classification using PCA, ICA, LDA and
support vector machines. Expert Systems with
Applications, 37(12), 8659–8666.
https://doi.org/10.1016/j.eswa.2010.06.065
Syafi’i, S. I., Wahyuningrum, R. T., & Muntasa, A.
(2016). Segmentasi Obyek Pada Citra Digital
Menggunakan Metode Otsu Thresholding.
Jurnal Informatika, 13(1), 1–8.
https://doi.org/10.9744/informatika.13.1.1-8
Whidhiasih, R. N., Wahanani, N. A., & Supriyanto,
S. (2013). Klasifikasi Buah Belimbing
Berdasarkan Citra Red-Green-Blue
Menggunakan Knn Dan Lda. Penelitian Ilmu
Komputer Sistem Embedded Dan Logic, 1(1),
29–35.
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10480
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative
Commons Attribution-NonCommercial 4.0 International License. 42
Analysis K-Nearest Neighbor Algorithm for
Improving Prediction Student Graduation Time
Rizki Muliono
Universitas Medan Area
Medan, Indonesia
Juanda Hakim Lubis
Universitas Medan Area
Medan, Indonesia
Nurul Khairina
Universitas Medan Area
Medan, Indonesia
Submitted: Jan 30, 2020
Accepted: Mar 11, 2020
Published: Apr 1, 2020
Abstract — Higher education plays a major role in improving the quality of education
in Indonesia. The BAN-PT institution established by the government has a standard of
higher education accreditation and study program accreditation. With the 4.0-based
accreditation instrument, it encourages university leaders to improve the quality and
quality of their education. One indicator that determines the accreditation of study
programs is the timely graduation of students. This study uses the K-Nearest Neighbor
algorithm to predict student graduation times. Students' GPA at the time of the seventh
semester will be used as training data, and data of students who graduate are used as
sample data. K-Nearest Neighbor works in accordance with the given sample data. The
results of prediction testing on 60 data for students of 2015-2016, obtained the highest
level of accuracy of 98.5% can be achieved when k = 3. Prediction results depend on the
pattern of data entered, the more samples and training data used, the calculation of the
K-Nearest Neighbor algorithm is also more accurate.
Keywords — prediction; graduation time; k-nearest neighbor
I. INTRODUCTION
In the Study Program and Higher Education
Accreditation Forms, the timely graduation of
undergraduate students is one component that has
influence (Novianti & Prasetyo, 2017). To get good
grades in accreditation, students are targeted to graduate
on time and achieve an average Semester Achievement
Index above 3.50.
According to the graduation data of the Faculty of
Engineering, Universitas Medan Area in recent years,
the average number of students who complete their
studies on time has not yet reached the target. Some
problems that often occur that cause students to
graduate on time, including low Semester Achievement
Index and GPA scores, economic factors,
environmental factors, and family.
Prediction of students will graduate on time or not
can be noticed since students sit in the seventh semester.
Semester Achievement Index sand the number of
credits will be a reference to predict the time students
graduate.
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10480
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative
Commons Attribution-NonCommercial 4.0 International License. 43
Data mining is one of the fields of computer science
that focuses on machine learning (Muliono, Muhathir,
Khairina, & Harahap, 2019) (Muliono, 2017). Data
mining is used to predict conditions based on data and
information (Tang, He, & Zhang, 2020) (Muliono &
Sembiring, 2019). The K-Nearest Neighbor method
uses data classification techniques that are divided into
clusters (Agrawal, 2019). Prediction results can be
calculated based on the distance closest to the sample
data (Gou et al., 2019) (Czumaj & Sohler, 2020). This
research will predict the graduation time of the students
with the K-Nearest Neighbor algorithm. As for some
previous studies related to this research are as follows:
The Research by (Prasetyo, Kusrini, & Arief, 2019)
uses the K-Nearest Neighbor algorithm to see the
interests and talents of students in the field of
Information Engineering. This choice of specialization
is done by Case Base Reasoning (CBR). The results
showed that this algorithm successfully predicted with
an accuracy rate of 95.98% at K = 7.
The Research by (Nikmatun & Waspada, 2019)
applies the K-Nearest Neighbor algorithm that refers to
Data Mining Knowledge Discovery in Database
(KDD). This study classifies courses that determine the
time students graduate. The research results obtained a
good prediction with an accuracy of 75.95%.
The research by (Hakim, Rizal, & Ratnasari, 2019)
uses the K-Nearest Neighbor algorithm and Roger S.
Pressman's waterfall method namely Communication,
Planning, Modeling, and Construction. The results
showed that the best accuracy was found in testing with
the Confusion Matrix, where the accuracy reached 98%.
The Research by (Rohman & Rochcham, 2019)
compares Neural Network, K-Nearest Neighbor and
Decision Tree algorithms in predicting student
graduation. The results showed that the highest
accuracy was found in the K-Nearest Neighbor
algorithm which reached 83.66%.
The research by (Purwanto, Kusrini, &
Sudarmawan, 2019) made a comparison of the C.45
algorithm and the K-Nearest Neighbor in predicting the
study period of students of Muhammadiyah University
in Purwokerto. The results showed that the highest
accuracy was found in the K-Nearest Neighbor
algorithm which reached 89.14%.
II. METHODOLOGY
K-Nearest Neighbor algorithm is a classification
method that can classify new data based on the distance
of the new data to the closest data/neighbors in data
learning (Atma & Setyanto, 2018) :
The training process is to start input: training data,
data transfer label, k, testing data.
a. For all testing data, calculate the distance to each
training data
b. Determine the training data k which is the closest
distance to the data
c. Testing
d. Check the label of this data
e. Determine the label with the most frequency
f. Enter the testing data to the class with the most
frequency
g. Stop
To calculate the distance between two points x and
y, you can use the Euclidean distance as follows (Wang
et al., 2019)
Which X1, 1 = 1, 2, is the category attribute, and n
1j – n2i represents the corresponding frequency. The closeness between the two cases can be calculated by finding the value of similarity as follows (Rahmatullah & Utami, 2019)
Description :
q: new case
s: cases that are in deviation
n: number of attributes in each case
i: individual attributes between 1 to n
f: similarity function I between cases T and S
wi: the weight is given to the i-th attribute
This similarity is expressed by 1 (similar) and 0 (not
similar), mathematically, it can be written:
𝒔 = 𝟏 𝒊𝒇 𝒙 = 𝒚𝟎 𝒊𝒇 𝒙 ≠ 𝒚
Giving weights for each attribute can be done by
following a few steps below:
1. Input the criteria value of each model (LCS)
2. Input the weights of each criterion (BBT)
3. Calculate normalization from weights (NK)
𝑉𝑎𝑙𝑢𝑒 = Σ 𝑁𝐾
𝑁
To test the accuracy of the predicted performance
measurement of the K-NN algorithm, it is performed by
(2)
(3)
(4)
(5)
(6)
(1)
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10480
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative
Commons Attribution-NonCommercial 4.0 International License. 44
comparing the results of the classification algorithm
prediction with the target value of the testing data
variable as the actual data. So logically, it can be
concluded that the performance of the algorithm is as
follows:
𝐴𝑐𝑐𝑢𝑟𝑎𝑐𝑦 = 𝑡ℎ𝑒 𝑝𝑟𝑒𝑑𝑖𝑐𝑡𝑒𝑑 𝑎𝑚𝑜𝑢𝑛𝑡 𝑖𝑠 𝑐𝑜𝑟𝑟𝑒𝑐𝑡
𝑡𝑜𝑡𝑎𝑙 𝑛𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑖𝑛𝑠𝑡𝑎𝑛𝑐𝑒𝑠 𝑥 100%
III. RESULT AND DISCUSSION
The research test was conducted on the data of 20
students of the Civil Engineering Study Program in the
seventh semester of 2015. Detailed research results can
be seen as follows:
Table 1. Accuracy Comparison of Predictive Predicate
Values with KNN and Real Data K = 5
No NPM Prediction
KNN Result Point
1 158110001 Correct Correct 1
2 158110002 Incorrect Incorrect 1
3 158110003 Incorrect Incorrect 1
4 158110005 Incorrect Incorrect 1
5 158110006 Correct Correct 1
6 158110007 Incorrect Incorrect 1
7 158110010 Correct Correct 1
8 158110012 Incorrect Incorrect 1
9 158110015 Incorrect Incorrect 1
10 158110017 Correct Correct 1
11 158110018 Incorrect Incorrect 1
12 158110020 Incorrect Correct 0
13 158110022 Correct Correct 1
14 158110023 Incorrect Incorrect 1
15 158110024 Incorrect Incorrect 1
16 158110025 Correct Correct 1
17 158110027 Correct Correct 1
18 158110028 Incorrect Incorrect 1
19 158110029 Incorrect Incorrect 1
20 158110030 Incorrect Incorrect 1
From the results of experiments conducted to see
the accuracy of the comparison of training data to the
results of algorithms found the results with timely
conclusions at K1 and K2 = 176, while K3-K5 = 197.
Table 2. Accuracy Levels
K Accuracy
Levels
Accuracy
Levels Conclusion
Confusion
Matrix
ROC
Curve
K1 88,0% 0.880 Good
Classification
K2 88,0% 0.880 Good
Classification
K3 98,5% 0,985 Excellent
Classification
K4 98,5% 0,985 Excellent
Classification
K5 98,5% 0,985 Excellent
Classification
The higher the K value, the better the accuracy
level of K-NN algorithm predictions on 2015 student
training data, the conclusion is from the K1-K5 trial
results of K-NN algorithm classification results in the
accuracy of student graduation prediction by comparing
the Rael scores and the prediction results can be
concluded as Excellent Classification
Next is to make a prediction on time for the 2016
data of the 2016 students' whip, which will be tested
from K1-K5.
Table 3. Predicted Results for 2016 Stock Data of
Civil Engineering Study Program
NPM SAI
1
SAI
2
SAI
3
SAI
4
SAI
5
SAI
6
SAI
7
SKS
Passed Prediction
168110003
3.25 3.29 3.82 3.53 3.53 3.82 3.53 136 Correct
168110005
3.61 3.47 2.04 1.75 2.67 2.55 3.37 122 Incorrect
168110009
3.38 3.13 2.55 3.37 3.05 3.61 3.29 132 Incorrect
16811
0011 3.39 3.13 3.47 2.95 2.67 3.82 2.55 132 Incorrect
168110012
3.61 3.03 3.37 3.42 3.47 3.76 2.67 136 Correct
168110016
3.29 3.71 2.88 3.05 3.53 2.88 3.05 132 Incorrect
168110017
3.24 3.13 3.61 3.29 2.55 3.05 3.53 132 Incorrect
168110022
3.71 3.29 3.05 3.53 3.05 3.82 3.53 136 Correct
16811
0026 3.61 3.47 2.04 1.75 2.67 2.55 3.37 122 Incorrect
168110028
3.61 3.47 2.04 1.75 2.67 2.55 3.37 122 Incorrect
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10480
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative
Commons Attribution-NonCommercial 4.0 International License. 45
The sample data of 200 data consists of the 2015
data stick and the data to be predicted is the 2016 data
canopy of 60 data with a ratio of 70% training data and
30% testing data. From the results of prediction
experiments on 2016 data, there are 60 data with timely
prediction results that can be seen in the following table:
Table 4. Predicted Results of 2016 Whamb Graduation
K Correct Incorrect
K1 25 35
K2 25 35
K3 16 44
K4 16 44
K5 16 44
IV. CONCLUSION AND SUGGESTION
A. Conclusion
The conclusions of this study are as follows:
1. In the case of predictions of 2015 student data on
the whip of K-NN algorithm the better level of K3
and so on is 98.5% from the previous K 88%
increased by 1.5%
2. In predictions, 60 of the 2016 canopy data shows
the condition of predicted data at K1 and K2 = 25
On-Time, while at K3 - K5 = 16 On Time.
3. The state of the predicted results depends on the
distribution of data patterns, the more data the
better the calculation of the K-NN algorithm
4. The more data the application transfers, the less
time it takes to process distance calculations.
B. Suggestion
It is hoped that the K-NN prediction application
based on web-based research results can be used by the
Faculty to assist monitoring as an EWS (Early Warning
System) for the academic development of students in
the Faculty of Engineering, Universitas Medan Area in
particular.
V. ACKNOWLEDGMENT
The researcher would like to thank the Universitas
Medan Area for funding the DIYA Research to
completion. Hopefully, this research is not only useful
for the Universitas Medan Area but also can be useful
for the development of science and society.
VI. REFERENCES
Agrawal, R. (2019). Integrated Parallel K-Nearest
Neighbor Algorithm. In Smart Intelligent
Computing and Applications (p. 479). Springer
Singapore. https://doi.org/10.1007/978-981-13-
1921-1
Atma, Y. D., & Setyanto, A. (2018). Perbandingan
Algoritma C4.5 dan K-NN dalam Identifikasi
Mahasiswa Berpotensi Drop Out. Metik Jurnal,
2(2), 31–37.
Czumaj, A., & Sohler, C. (2020). Sublinear time
approximation of the cost of a metric k -nearest.
In Society for Industrial and Applied Mathematics
(pp. 2973–2992).
Gou, J., Ma, H., Ou, W., Zeng, S., Rao, Y., & Yang, H.
(2019). A Generalized Mean Distance-Based K-
Nearest Neighbor Classifier. Expert Systems with
Applications, 115, 3–24.
https://doi.org/10.1016/j.eswa.2018.08.021
Hakim, L. A. R., Rizal, A. A., & Ratnasari, D. (2019).
Aplikasi Prediksi Kelulusan Mahasiswa Berbasis
K-Nearest Neighbor (K-NN). JTIM : Jurnal
Teknologi Informasi Dan Multimedia, 1(1), 30–
36. https://doi.org/10.35746/jtim.v1i1.11
Muliono, R. (2017). Implementasi Algoritma Apriori
Pada Data Benchmark Kosarak Dan Mushrooms.
Journal of Informatics and Telecommunication
Engineering, 1(1), 34–41.
Muliono, R., Muhathir, Khairina, N., & Harahap, M. K.
(2019). Analysis of Frequent Itemsets Mining
Algorithm Againts Models of Different Datasets.
In 1st International Conference of SNIKOM 2018
(pp. 1–5). https://doi.org/10.1088/1742-
6596/1361/1/012036
Muliono, R., & Sembiring, Z. (2019). Data Mining
Clustering Menggunakan Algoritma K-Means
Untuk Klasterisasi Tingkat Tridarma Pengajaran
Dosen. CESS (Journal Of Computer Engineering,
System And Science), 4(2), 272–279.
Nikmatun, I. A., & Waspada, I. (2019). Implementasi
Data Mining untuk Klasifikasi Masa Studi
Mahasiswa Menggunakan Algoritma K-Nearest
Neighbor. Jurnal Simetris, 10(2), 421–432.
Novianti, A. G., & Prasetyo, D. (2017). Penerapan
Algoritma K-Nearest Neighbor (K-NN) untuk
Prediksi Waktu Kelulusan Mahasiswa. In
Seminar Nasional APTIKOM(SEMNASTIKOM)
(pp. 108–113).
Prasetyo, A., Kusrini, & Arief, M. R. (2019). Penerapan
Algoritma K Nearest Neighbor untuk
Rekomendasi Minat Konsentrasi di Program
Studi Teknik Informatika Universtias PGRI
Yogyakarta. Informasi Interaktif, 4(1), 1–6.
Purwanto, E., Kusrini, & Sudarmawan. (2019). Prediksi
Kelulusan Tepat Waktu Menggunakan Metode
C4 . 5 DAN K - NN (Studi Kasus : Mahasiswa
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10480
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative
Commons Attribution-NonCommercial 4.0 International License. 46
Program Studi S1 Ilmu Farmasi , Fakultas
Universitas Muhammadiyah Purwokerto ).
TECHNO, 20(2), 131–142.
Rahmatullah, S., & Utami, E. (2019). Prediksi Tingkat
Kelulusan Tepat Waktu dengan Metode Naive
Bayes dan K-Nearest Neighbor. Jurnal Informasi
Dan Komputer, 7(1), 7–16.
Rohman, A., & Rochcham, M. (2019). Komparasi
Metode Klasifikasi Data Mining untuk Prediksi
Kelulusan Mahasiswa. Jurnal Neo Teknika, 5(1),
23–31.
Tang, B., He, H., & Zhang, S. (2020). MCENN: A
Variant of Extended Nearest Neighbor Method
for Pattern Recognition. Pattern Recognition
Letters, 1–10.
https://doi.org/10.1016/j.patrec.2020.01.015
Wang, Y., Wang, R., Li, D., Adu-Gyamfi, D., Tian, K.,
& Zhu, Y. (2019). Improved Handwritten Digit
Recognition using Quantum K-Nearest Neighbor
Algorithm. International Journal of Theoretical
Physics, 58(7), 2331–2340.
https://doi.org/10.1007/s10773-019-04124-5
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10502
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative
Commons Attribution-NonCommercial 4.0 International License. 47
Apriori Algorithm on Car Rental Analysis with
The Most Popular Brands
Leo Fernando Panjaitan
STMIK Nusa Mandiri
Jakarta, Indonesia
Yopi Handrianto
Universitas Bina Sarana Informatika
Jakarta, Indonesia
Achmad Nurhadi
Universitas Bina Sarana Informatika
Jakarta, Indonesia
Submitted: Feb 26, 2020
Accepted: Mar 17, 2020
Published: Apr 1, 2020
Abstract— Nowadays, vehicle rental has become a common function for companies that
have busy operational activities. Every company in carrying out operational activities
requires a vehicle that is always there when needed. PT. Agung Solusi Trans is a vehicle
rental company that rents various vehicle brands commonly used by customers to rent
vehicles. In addition, PT. Agung Solusi Trans is also difficult to get updated information
regarding the level of sales per period. Therefore, we need a decision support system
and a method that can be used to design a business strategy that can provide an efficient
and effective information, namely data mining using the a priori algorithm association
method. The researcher specializes in taking only vehicle types as research material by
selecting fifteen brands, including Agya, Yaris, Sienta, Calya, Avanza, Innova, Rush,
Vios, Altis, Camry, Fortuner, Alphard, Hi Ace, Voxy, and Hilux. In analyzing the data,
the writer uses a priori algorithm calculation by testing the hypothesis of two variables
between the value of support and the value of confidence. After that, apriori algorithm
is calculated using Tanagra. Based on the analysis done by the author, that the brands
most sought after by customers are Calya, Avanza, Hilux. From these results can be used
by PT. Agung Solusi Trans to prepare vehicle brands that are widely leased by customers
and increase brand inventory.
Keywords— Apriori Algorithms, Analysis, Brands
I. INTRODUCTION
At this time the growth of business in the field of
vehicle rental services is being looked after by people
who have the need to overcome even odd numbers on
the company's operational vehicles. With the
existence of an even odd system created by the
government, it is not surprising that many newcomer
rental services have sprung up in the world of vehicle
rental
One of the means of transportation that has a good
function and with a lot of transport capacity, as well
as easy and inexpensive to carry and rent is a car. In
its development, the car rental business has become a
very profitable business or business. Because at this
time people are more pleased use the car to travel far
and in the distance that is being traveled. For most
people traveling by car will be more enjoyable and
comfortable in the trip and can bring many family
members, friends, or friends who take part in the trip.
So that the pleasure in traveling will be easier to
obtain when compared to using other means of
transportation (Septavia et al., 2014).
Based on the above background, this study discusses data mining by classifying vehicle brands that are often rented by customers using a priori
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10502
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative
Commons Attribution-NonCommercial 4.0 International License. 48
algorithm association rules to determine the pattern of item and itemset combination of vehicle rental sales data transactions at PT. Agung Solusi Trans for one year..
II. LITERATURE REVIEW
A. Understanding Data Mining
According to (Badrul, 2015) "Data mining is a technique which is a combination of methods of continuous data analysis with algorithms for processing large data.”.
According to (Handrianto & Farhan, 2019) “Data Mining is a process that uses statistical, mathematical, artificial intelligence and machine learning techniques to extract and identify useful information and related knowledge from various large databases”.
According to (Irfiani, 2019), “Data mining is a method by extracting large amounts of information, in this way helping companies focus on important information in data warehouses”.
According to (Anas, 2016), "Data mining algorithms mostly come from the form of developing algorithms in various fields of machine learning, statistics, artificial intelligence and artificial neural networks. Because it is not designed to handle data in very large sizes, while data mining in question is useful for handling data of such size, then one direction of research in the field of data mining is to develop these algorithms so that they can handle very large data sizes”.
There are two reasons why data mining needs to be used, namely:
a. Finding patterns in the data can be disappointing for inexperienced decision makers due to the fact that potential patterns in the data are often invisible.
b. The amount of data is too large to manually analyze.
According to (Santoso et al., 2016), data mining has its own characteristics, among others:
a. Data mining must relate to the discovery of something that is hidden and has certain data patterns that were not previously known.
b. Data mining usually uses very large data
c. Data mining is used to make a critical decision, especially in determining a strategy. Based on some of the opinions mentioned above it can be concluded that data mining is the process of finding the results of large amounts of data stored in databases, data warehouses, sales data or other storage media. As a
series of processes, data mining is divided into several process stages. These stages are interactive, the user is directly involved or through the mediation of a knowledge base (basic knowledge).
There are six data mining stages, including (Halimi et al., 2019) :
a. Data cleaning is the process of removing noise and inconsistent or irrelevant data.
b. Data Integration is the merging of data from various databases into one new database.
c. Data Selection is the data that is in the database is often not all used, therefore only the data that is suitable for analysis will be taken from the database.
d. Data transformation is data that is changed or merged into a format suitable for processing in data mining.
e. Mining process is the main process when the method is applied to find valuable and hidden knowledge from data.
f. Pattern Evaluation (Pattern Evaluation) is used to identify interesting patterns into the knowledge based found.
g. Knowledge presentation is the visualization and presentation of knowledge about the method used to obtain the knowledge obtained by the user.
B. Definition of Association Rule
According to (Kanti & Indrajit, 2017) states that, "analysis of association rules in data mining is a data mining technique for finding association rules between a combination of items". The association rule analysis is known as one of the data mining techniques that is the basis of various other data mining techniques. The analysis phase of association rules will produce an efficient algorithm using high frequency patterns. This association rule aims to look for patterns that often occur among many transactions, where each transaction consists of several items.
According to Kusrini in (Listriani et al., 2016) stated that "The basic methodology of association analysis consists of two, namely: analysis of high frequency patterns and the formation of associative rules".
a. High Frequency Pattern Analysis
This stage looks for combinations of items that meet the minimum requirements of the support value in the database. The value of an item's support is obtained by the following formula.
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10502
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative
Commons Attribution-NonCommercial 4.0 International License. 49
𝑺𝒖𝒑𝒑𝒐𝒓𝒕 (𝑨) =𝐓𝐨𝐭𝐚𝐥 𝐓𝐫𝐚𝐧𝐬𝐚𝐜𝐭𝐢𝐨𝐧 𝐂𝐨𝐧𝐭𝐚𝐢𝐧𝐢𝐧𝐠 𝐀
𝐓𝐨𝐭𝐚𝐥 𝐓𝐫𝐚𝐧𝐬𝐚𝐜𝐭𝐢𝐨𝐧∗ 𝟏𝟎𝟎 %
While the support values of the 2 items are obtained from the following formula 2.
𝑺𝒖𝒑𝒑𝒐𝒓𝒕(𝐀, 𝐁)∑ 𝐓𝐫𝐚𝐧𝐬𝐚𝐜𝐭𝐢𝐨𝐧 𝐂𝐨𝐧𝐭𝐚𝐢𝐧𝐢𝐧𝐠 𝐀 𝐚𝐧𝐝 𝐁
∑ 𝐓𝐫𝐚𝐧𝐬𝐚𝐜𝐭𝐢𝐨𝐧∗ 𝟏𝟎𝟎%
b. Establishment of Associative Rules
The formation of the association rules is done after all high frequency patterns have been found, then the association rules are found to meet the minimum requirements for confidence by calculating the associative rule confidence "if A then B". The confidence value of the rule "if A then B" is obtained from the following formula:
𝒄𝒐𝒏𝒇𝒊𝒅𝒆𝒏𝒄𝒆(𝐀, 𝐁)
=∑ 𝐓𝐎𝐓𝐀𝐋 𝐓𝐑𝐀𝐍𝐒𝐀𝐂𝐓𝐈𝐎𝐍𝐂𝐎𝐍𝐓𝐀𝐈𝐍𝐈𝐍𝐆 𝐀 𝐚𝐧𝐝 𝐁
∑ 𝐓𝐎𝐓𝐀𝐋 𝐓𝐑𝐀𝐍𝐒𝐀𝐂𝐓𝐈𝐎𝐍 𝐀∗ 𝟏𝟎𝟎%
C. Definition of Apriori Algorithm
According to (Badrul, 2016), "Apriori algorithm is a basic algorithm proposed by Agrawal & Srikant in 1994 to determine frequent itemsets as a rule of boolean association. A priori algorithm including the type of association rules in data mining”.
Broadly speaking, the work of a priori algorithms is:
a. Formation of itemset candidates, k-itemset candidates are formed from a combination (k-1) -itemset obtained from the previous iteration. One feature of the Apriori algorithm is the pruning of k-itemset candidates whose subsets containing k-1 items are not included in the high frequency pattern with k-1 length.
b. Calculation of support for each k-itemset candidate. Support from each k-itemset candidate is obtained by scanning the database to count the number of transactions containing all items in the k-itemset candidate. This is also a feature of the a priori algorithm which requires the calculation by scanning the entire database of the longest k-itemset.
c. Set high frequency pattern. High frequency patterns containing k items or k-itemset are determined from candidate k-itemset whose support is greater than the minimum support. If no new high frequency pattern is obtained, the whole process is stopped. If not, then k plus one and return to part 1.
D. Definition of Tanagra
According to (Badrul, 2016), "Tanagra is a free and useful data mining software for academic purposes and can be taken from several data mining
methods in the form of data exploration analysis, statistical learning, machine learning and database areas". Tanagra software is said to be free, because the software is open source where every researcher can access to the source code, and add his own algorithm, as long as the researcher agrees and complies with the software distribution license.
II. PROPOSED METHOD
A. Research Stages
According to (Badrul, 2016), "Tanagra is a free and useful data mining software for academic purposes and can be taken from several data mining methods in the form of data exploration analysis, statistical learning, machine learning and database areas". Without the software is said to be free, because the software is open source where every researcher can access to the source code, and add his own algorithm, as long as the researcher agrees and complies with the software distribution license:
Source : Sikumbang (2018)
Figure 1. Research Stages
B. Data Collection Methods, Population and Research Samples
Method of collecting data
The research data collection was carried out in the following way:
1. Observation or observation
Observing information relating to the Apriori Algorithm Method at PT. Agung Solusi Trans. The results of direct observation are then recorded by the author so that it can be seen the application that can be done with the method.
2. Interview
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10502
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative
Commons Attribution-NonCommercial 4.0 International License. 50
The author conducted a direct interview with the marketing manager of PT. Agung Solusi Trans who knows product inventory data and vehicle rental sales transaction data that occurred in the company for a year.
3. Literature Review
Study the journal literature and reference books supporting the theory contained in the library.
Study population and sample
“Population is a generalization area in the form of subjects or objects studied and has certain qualities and characteristics determined by researchers to be studied, then conclusions are drawn ". (Sugiyono, 2015)
Whereas "The sample is part or acts as a representative of the population so that the results of research successfully obtained from the sample can be generalized to the population". (Sugiyono, 2015)
The population in this study is sales data seen from customers or customers who do all vehicle rental transactions at PT. Agung Solusi Trans. The object of the sample taken in this study, namely: vehicle rental data categorized by brand and month period in 2018. The sampling technique is a sampling technique. The technique used in this research is systematic nonprobability sampling where this sampling technique does not provide the same opportunity or opportunity for each population element selected to be sampled and taken based on the order of the population elements that have been given a sequence number.
Table 1. Vehicle Rental Data Samples
Month Item set
1 Rush, Avanza, Hilux
2 Innova, Avanza, Rush
3 Vios, Camry, Innova
4 Hilux, Innova, Avanza
5 Avanza, Vios, Calya
6 Avanza, Vios, Innova
7 Alphard, Rush, Vios
8 Avanza, Voxy, Sienta
9 Innova, Hi Ace, Rush
10 Innova, Fortuner, Yaris
11 Avanza, Altis, Camry
12 Avanza, Alphard, Calya
Source: Research Results (2019)
C. Data Analysis Method
The study was conducted on the selection of vehicle brands that are most interested in customers with the acquisition of vehicle rental data collection at PT. Agung Solusi Trans. In analyzing the data, the writer uses a priori algorithm calculation by testing the hypothesis of two variables between the value of support and the value of confidence. After that, a priori algorithm is calculated using Tanagra which functions as a matching of the results obtained in the previous calculation.
1. Apriori Algorithm Calculation
A. High Frequency Pattern Analysis
At this stage, the writer looks for a combination of items that meet the minimum requirements of the support value in the database. The value of an item's support is obtained by the following formula:
𝑺𝒖𝒑𝒑𝒐𝒓𝒕 (𝑨) =𝐓𝐨𝐭𝐚𝐥 𝐓𝐫𝐚𝐧𝐬𝐚𝐜𝐭𝐢𝐨𝐧 𝐂𝐨𝐧𝐭𝐚𝐢𝐧𝐢𝐧𝐠 𝐀
𝐓𝐨𝐭𝐚𝐥 𝐓𝐫𝐚𝐧𝐬𝐚𝐜𝐭𝐢𝐨𝐧∗ 𝟏𝟎𝟎 %
While the support value of 2 items is obtained from the following formula:
Support (A,B) = P(A∩B)
𝑺𝒖𝒑𝒑𝒐𝒓𝒕(𝐀, 𝐁)∑ 𝐓𝐫𝐚𝐧𝐬𝐚𝐜𝐭𝐢𝐨𝐧 𝐂𝐨𝐧𝐭𝐚𝐢𝐧𝐢𝐧𝐠 𝐀 𝐚𝐧𝐝 𝐁
∑ 𝐓𝐫𝐚𝐧𝐬𝐚𝐜𝐭𝐢𝐨𝐧∗ 𝟏𝟎𝟎%
And for the support value of 3 items obtained from the following formula:
𝑺𝒖𝒑𝒑𝒐𝒓𝒕(𝐀, 𝐁, 𝐂)∑ 𝐓𝐫𝐚𝐧𝐬𝐚𝐜𝐭𝐢𝐨𝐧 𝐂𝐨𝐧𝐭𝐚𝐢𝐧𝐢𝐧𝐠 𝐀, 𝐁 𝐝𝐚𝐧 𝐂
∑ 𝐓𝐫𝐚𝐧𝐬𝐚𝐜𝐭𝐢𝐨𝐧∗ 𝟏𝟎𝟎%
B. Formation of Association Rules
After all high frequency patterns have been found and the iteration has stopped, then the association rules are found to meet the minimum requirements for confidence by calculating the confidence of associative rules. A → B.
The confidence value of the rules A → B is obtained from the following formula :
𝒄𝒐𝒏𝒇𝒊𝒅𝒆𝒏𝒄𝒆(𝐀, 𝐁) =∑ 𝐓𝐨𝐭𝐚𝐥 𝐓𝐫𝐚𝐧𝐬𝐚𝐜𝐭𝐢𝐨𝐧 𝐂𝐨𝐧𝐭𝐚𝐢𝐧𝐢𝐧𝐠 𝐀 𝐚𝐧𝐝 𝐁
∑ 𝐓𝐎𝐓𝐀𝐋 𝐓𝐑𝐀𝐍𝐒𝐀𝐂𝐓𝐈𝐎𝐍 𝐀∗ 𝟏𝟎𝟎%
2. Apriori Algorithm Calculation with Tanagra
Apriori algorithm on Tanagra can be formed with predetermined steps. The steps in forming the algorithm are as follows:
A. Support Algorithm
In determining the support algorithm there are inputs, outputs and processes. The following are the results of input, output and data processing:
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10502
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative
Commons Attribution-NonCommercial 4.0 International License. 51
Figure 2. Data Input Results
Figure 3. Output Data Results
Figure 4. Itemset Search process
Figure 5. Support Itemset Display Results
B. Confidence Algorithm
In determining the support algorithm there are inputs, outputs and processes. The following are the results of input, output and data processing:
Figure 6. Input Looking for Confidence
Figure 7. Results for Confidence
III. RESULT AND DISCUSSION
Highest Frequency Pattern Analysis
A. 1 Itemset combination
With a minimum support value of 30%, the support value of 1 item is obtained by the following formula:
𝑺𝒖𝒑𝒑𝒐𝒓𝒕(𝐀, 𝐁) =∑ 𝐓𝐫𝐚𝐧𝐬𝐚𝐜𝐭𝐢𝐨𝐧 𝐂𝐨𝐧𝐭𝐚𝐢𝐧𝐢𝐧𝐠 𝐀
∑ 𝐓𝐫𝐚𝐧𝐬𝐚𝐜𝐭𝐢𝐨𝐧∗ 𝟏𝟎𝟎%
Following is the calculation of the formation of 1 itemset:
𝑆(Rush) =∑ Rush
∑ 12=
∑ 4
∑ 12∗ 100% = 33.33%
𝑆(Vios) =∑ Vios
∑ 12=
∑ 4
∑ 12∗ 100% = 33.33%
𝑆(Avanza) =∑ Avanza
∑ 12=
∑ 8
∑ 12∗ 100% = 66.70%
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10502
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative
Commons Attribution-NonCommercial 4.0 International License. 52
𝑆(Innova) =∑ Innova
∑ 12=
∑ 6
∑ 12∗ 100% = 50.00%
The support value of 1 item that has been described can be seen in the table below:
Table 2. Support 1 Itemset
Itemset Support Support (%)
Agya 0/12 0.0%
Yaris 1/12 8.3%
Sienta 1/12 8.3%
Calya 2/12 16.7%
Avanza 8/12 66.7%
Innova 6/12 50.0%
Rush 4/12 33.3%
Vios 4/12 33.3%
Altis 1/12 8.3%
Camry 2/12 16.7%
Fortuner 1/12 8.3%
Alphard 2/12 16.7%
Hi Ace 1/12 8.3%
Voxy 1/12 8.3%
Hilux 2/12 16.7%
Land Cruiser 0/12 0.0%
Source : Research result (2019)
With a minimum support of 30%, then the combination of 1 itemset that does not meet the minimum support will be removed in the following table:
Table 3. Minimum Support 1 Itemset
Itemset Support Support(%)
Rush 4/12 33.3%
Vios 4/12 33.3%
Innova 6/12 50.0%
Avanza 8/12 66.7%
Source : Research result (2019)
B. 2 itemset combination
Formation of Combinations The support values of 2 items are obtained by the following formula:
𝑺𝒖𝒑𝒑𝒐𝒓𝒕(𝐀, 𝐁)∑ 𝐓𝐫𝐚𝐧𝐬𝐚𝐜𝐭𝐢𝐨𝐧 𝐂𝐨𝐧𝐭𝐚𝐢𝐧𝐢𝐧𝐠 𝐀 𝐚𝐧𝐝 𝐁
∑ 𝐓𝐫𝐚𝐧𝐬𝐚𝐜𝐭𝐢𝐨𝐧∗ 𝟏𝟎𝟎%
The following is a calculation of 2 item sets:
𝑆(Yaris, Innova) =∑ Yaris, Innova
∑ 12=
1
12∗ 100%
= 8.33%
𝑆(Yaris, Fortuner) =∑ Yaris,Fortuner
∑ 12=
1
12∗ 100% =
8.33%
𝑆(Sienta, Avanza) =∑ Sienta, Avanza
∑ 12=
1
12∗ 100%
= 8.33%
𝑆(Sienta, Voxy) =∑ Sienta,Voxy
∑ 12=
1
12∗ 100% = 8.33%
𝑆(Calya, Avanza) =∑ Calya, Avanza
∑ 12=
2
12∗ 100%
= 16.67%
𝑆(Calya, Vios) =∑ Calya, Vios
∑ 12=
1
12∗ 100% = 8.33%
𝑆(Avanza, Innova) =∑ Avanza, Innova
∑ 12=
3
12∗ 100%
= 25.00%
𝑆(Avanza, Rush) =∑ Avanza, Rush
∑ 12=
2
12∗ 100%
= 16.67%
𝑆(Avanza, Vios) =∑ Avanza, Vios
∑ 12=
2
12∗ 100%
= 16.67%
𝑆(Avanza, Altis) =∑ Avanza, Altis
∑ 12=
1
12∗ 100%
= 8.33%
𝑆(Avanza, Camry) =∑ Avanza, Camry
∑ 12=
1
12∗ 100%
= 8.33%
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10502
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative
Commons Attribution-NonCommercial 4.0 International License. 53
𝑆(Avanza, Alphard) =∑ Avanza, Alphard
∑ 12=
1
12∗ 100%
= 8.33%
𝑆(Avanza, Voxy) =∑ Avanza, Voxy
∑ 12=
1
12∗ 100%
= 8.33%
𝑆(Avanza, Hilux) =∑ Avanza, Hilux
∑ 12=
2
12∗ 100%
= 16.67%
𝑆(Innova, Rush) =∑ Innova, Rush
∑ 12=
2
12∗ 100%
= 16.67%
𝑆(Innova, Vios) =∑ Innova, Vios
∑ 12=
2
12∗ 100%
= 16.67%
𝑆(Innova, Fortuner) =∑ Innova, Fortuner
∑ 12
=1
12∗ 100% = 8.33%
𝑆(Rush, Vios) =∑ Rush, Vios
∑ 12=
1
12∗ 100% = 8.33%
𝑆(Rush, Alphard) =∑ Rush, Alphard
∑ 12=
1
12∗ 100%
= 8.33%
𝑆(Rush, Hi Ace) =∑ Rush, Hi Ace
∑ 12=
1
12∗ 100%
= 8.33%
𝑆(Rush, Hilux) =∑ Rush, Hilux
∑ 12=
1
12∗ 100% = 8.33%
𝑆(Vios, Camry) =∑ Vios, Camry
∑ 12=
1
12∗ 100% = 8.33%
𝑆(Vios, Alphard) =∑ Vios, Alphard
∑ 12=
1
12∗ 100%
= 8.33%
𝑆(Altis, Camry) =∑ Altis, Camry
∑ 12=
1
12∗ 100%
= 8.33%
The support value of the 2 items that have been obtained from the description can be seen in the table below :
Table 4. Support 2 Itemset
Item set Support Support (%)
Yaris, Innova 1/12 8.33%
Yaris, Fortuner 1/12 8.33%
Sienta, Avanza 1/12 8.33%
Sienta, Voxy 1/12 8.33%
Calya, Avanza 2/12 16.67%
Calya, Vios 1/12 8.33%
Avanza, Innova 3/12 25.00%
Avanza, Rush 2/12 16.67%
Avanza, Vios 2/12 16.67%
Avanza, Altis 1/12 8.33%
Avanza, Camry 1/12 8.33%
Avanza, Alphard 1/12 8.33%
Avanza, Voxy 1/12 8.33%
Avanza, Hilux 2/12 16.67%
Innova, Rush 2/12 16.67%
Innova, Vios 2/12 16.67%
Innova, Fortuner 1/12 8.33%
Rush, Vios 1/12 8.33%
Rush, Alphard 1/12 8.33%
Rush, Hi Ace 1/12 8.33%
Rush, Hilux 1/12 8.33%
Vios, Camry 1/12 8.33%
Vios, Alphard 1/12 8.33%
Altis, Camry 1/12 8.33%
Source : Research result (2019)
With a minimum support of 30%, then the combination of 2 itemsset all of them not meeting the minimum support will be removed in the following table:
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10502
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative
Commons Attribution-NonCommercial 4.0 International License. 54
Table 5. Minimum Support 2 Itemset
Item set Support Support (%)
Calya, Avanza 2/12 16.67%
Avanza, Innova 3/12 25.00%
Avanza, Rush 2/12 16.67%
Avanza, Vios 2/12 16.67%
Avanza, Hilux 2/12 16.67%
Innova, Rush 2/12 16.67%
Innova, Vios 2/12 16.67%
Source : Research result (2019)
C. 3 itemset combination
Formation Combination The support value of 3 items is obtained by the following formula:
𝑺𝒖𝒑𝒑𝒐𝒓𝒕(𝐀, 𝐁, 𝐂)∑ 𝐓𝐫𝐚𝐧𝐬𝐚𝐜𝐭𝐢𝐨𝐧 𝐂𝐨𝐧𝐭𝐚𝐢𝐧𝐢𝐧𝐠 𝐀, 𝐁 𝐚𝐧𝐝 𝐂
∑ 𝐓𝐫𝐚𝐧𝐬𝐚𝐜𝐭𝐢𝐨𝐧∗ 𝟏𝟎𝟎%
The following is a calculation of 3 item sets :
𝑆(Rush, Avanza, Hilux ) =∑ Rush, Avanza, Hilux
∑ 12=
1
12∗ 100%
= 8.33%
The support value of the 3 items that have been obtained can be seen in the table below:
Table 6. Itemset Combination
Itemset Support Support(%)
Rush, Avanza, Hilux 1/12 8.33%
Source : Research result (2019)
Because of the formation of 3 itemset none meet the 16% support value, then the combination of 2 itemset that meets the formation of the Association.
Formation of Association Rules
After all the high frequency patterns have been found, then the association rules are found that meet the minimum requirements for confidence by calculating the confidence of associative rules A → B. Minimum Confidence = 60% Confidence value of the rules A → B is obtained by the following formula:
𝑪𝒐𝒏𝒇𝒊𝒅𝒆𝒏𝒄𝒆(𝐀, 𝐁)
=∑ 𝐓𝐎𝐓𝐀𝐋 𝐓𝐑𝐀𝐍𝐒𝐀𝐂𝐓𝐈𝐎𝐍 𝐂𝐎𝐍𝐓𝐀𝐈𝐍𝐈𝐍𝐆 𝐀 𝐚𝐧𝐝 𝐁
∑ 𝐓𝐎𝐓𝐀𝐋 𝐓𝐑𝐀𝐍𝐒𝐀𝐂𝐓𝐈𝐎𝐍 𝐀∗ 𝟏𝟎𝟎%
The following is a calculation of 3 item sets:
𝐶(Calya, Avanza) =∑ Calya, Avanza
∑ Calya=
2
2∗ 100%
= 100%
𝐶(Hilux, Avanza) =Hilux, Avanza
∑ Hilux=
2
2∗ 100% = 100%
From the combination of 2 itemsset that has been found, it can be seen the amount of support value, and the confidence of the candidate association rules as shown in the table below:
Table 7. Support Confidence
Rules Confidence
If you buy Calya, you will buy Avanza 2/2 100%
If you buy Hilux, you will buy Avanza 2/2 100%
Source : Research result (2019)
Final Association Rules
Final association rules are based on a minimum of 16% support and 60% confidence that has been determined, can be seen in the table below:
Table 8. Final Association
Rules Support Confidence
If you buy Calya, you will
buy Avanza 16.67% 100%
If you buy Hilux, you will
buy Avanza 16.67% 100%
Source : Research result (2019)
So based on the rules of the final association known from the transaction above, that the carrier that is most in demand by climbers is Calya, Avanza, Hilux. The results obtained can be seen in the diagram below:
Figure 8. Diagram of the Result of Forming Final Association Rules
In this section, the researcher will explain the results of the research obtained. Researchers can also use images, tables and curves to explain the results of the study. After finishing explaining the results of the study, researchers may give a simple discussion related to the results of the research trials.
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10502
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative
Commons Attribution-NonCommercial 4.0 International License. 55
IV. CONCLUSION AND SUGGESTION
Based on the discussion that has been carried out by the author using the Apriori Algorithm and testing with the Tanagra application, we get the following conclusions::
1. Apriori Algorithm Method has been successfully applied in analyzing the brand of vehicles rented in a company based on vehicle rental data transactions that occurred in 2018 within a year. From the results of the calculation of the analysis, the author can know that the most leased vehicle brands by customers at PT. Agung Solusi Trans are Avanza (66.7%), Innova (50.00%), Rush (33.33%) and Vios (33.33%). However, Agya (0%) and Land Cruiser (0%) are considered by customers as brands that are less desirable or rarely rented.
2. PT. Agung Solusi Trans as a provider of brand vehicle rental services makes the calculation results with the Apriori Algorithm method to obtain very valuable information in decision makers in the company. So that PT. Agung Solusi Trans can purchase vehicles and stock with brands that are preferred by tenants and increase the inventory of those brands that will be needed in the future. In addition, in implementing and improving the right marketing strategies in order to bring benefits to the PT. Agung Solusi Trans is done by making discounts on vehicle rentals with certain brands that are rarely rented to attract tenants.
3. The results obtained from the calculation of the Apriori Algorithm analysis method, can be used as a reference for tenants in choosing a vehicle brand that suits the tenant's operational needs.
4. Apriori Algorithm testing using the Tanagra application shows that the results obtained are the same as doing Apriori Algorithm calculations manually. This can be seen through the results of calculations in the value of support and confidence.
V. REFERENCES
Anas, A. (2016). Analisa Algoritma Apriori Untuk
Mendapatkan Pola Peminjaman Buku
Perpustakaan Smpn 3 Batanghari. Jurnal
Ilmiah Media SISFO, 10(2), 628–641.
http://ejournal.stikomdb.ac.id/index.php/media
sisfo/article/view/233/220
Badrul, M. (2015). Prediksi Hasil Pemilu Legislatif
Dengan Menggunakan Algoritma K-Nearest
Neighbor. Jurnal Pilar Nusa Mandiri, XI(2),
152–160.
http://ejournal.nusamandiri.ac.id/index.php/pil
ar/article/view/424/374
Badrul, M. (2016). Algoritma Asosiasi Dengan
Algoritma Apriori Untuk Analisa Data
Penjualan. None, 12(2), 121–129.
https://media.neliti.com/media/publications/22
7549-algoritma-asosiasi-dengan-algoritma-
apri-f4245cc8.pdf
Halimi, I., Azhar, Y., & Marthasari, G. I. (2019).
Prediksi Harga Emas Menggunakan Univariate
Convolutional Neural Network. Repositor,
1(2), 105–116.
https://doi.org/10.22219/repositor.v1i2.612
Handrianto, Y., & Farhan, M. (2019). C.45 Algorithm
for Classification of Causes of Landslides.
SinkrOn, 4(1), 120.
https://doi.org/10.33395/sinkron.v4i1.10154
Irfiani, E. (2019). Application of Apriori Algorithms
to Determine Associations in Outdoor Sports
Equipment Stores. SinkrOn, 3(2), 218.
https://doi.org/10.33395/sinkron.v3i2.10089
Kanti, S., & Indrajit, R. E. (2017). Implementasi Data
Mining Penjualan Handphone Oppo Store Sdc
Tanggerang Dengan Algoritma Appriori.
Implementasi Data Mining Penjualan
Handphone Oppo Store Sdc Tanggerang
Dengan Algoritma Appriori, November, 1–2.
Listriani, D., Setyaningrum, A. H., & Eka, F. (2016).
Penerapan Metode Asosiasi Menggunakan
Algoritma Apriori Pada Aplikasi Analisa Pola
Belanja Konsumen (Studi Kasus Toko Buku
Gramedia Bintaro). Jurnal Teknik Informatika,
9(2), 120–127.
https://doi.org/10.15408/jti.v9i2.5602
Santoso, H., Hariyadi, I. P., & Prayitno. (2016). Data
Mining Analisa Pola Pembelian Produk
Dengan Menggunakan Metode Algoritma
Apriori. Teknik Informatika, 1, 19–24.
http://ejournal.stikom-
db.ac.id/index.php/mediasisfo/article/view/233
/220
Septavia, I., Gunadhi, R. E., & Kurniawati, R. (2014).
Sistem Informasi Penyewaan Mobil Berbasis
Web di GTS Mobil Bandung. 13(1999), 10–24.
https://jurnal.sttgarut.ac.id/index.php/algoritm
a/article/view/247/411
Sugiyono. (2015). Metode penelitian pendekatan
kuantitatif, kualitatif dan R&DNo Title. In
Metode penelitian pendekatan kuantitatif,
kualitatif dan R&D (pp. 80–84). Alfabeta.
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10517
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative
Commons Attribution-NonCommercial 4.0 International License. 56
Pythagoras Tree Applied For Determined
Instagram Usage Habit Decision
Erlin Windia Ambarsari
Universitas Indraprasta PGRI
Jakarta, Indonesia
Herlinda
Universitas Indraprasta PGRI
Jakarta, Indonesia
Submitted: Mar 1, 2020
Accepted: Mar 15, 2020
Published: Apr 1, 2020
Abstract— In previous studies, Pythagoras Tree constructed using the Regression
Method, namely ID3 of Standard Deviation Reduction (SDR). The study using SDR for
Classification, which uses the Coefficient of Variant (CV) rather than Information Gain.
Data obtained from previous research about Instagram Usage Habit. The result of the
study that SDR is useful for the classification method for constructing Pythagoras Tree.
However, the target attribute is must use a numerical variable to gain the Standard
Deviation and the Mean. Empty data does not affect calculations. Although instances
must be discarded, thereby reducing the amount of data. For the case of the Instagram
Usage Habit itself, not getting the right pattern to deciding due to the data obtained is
less. The reason is Instagram Usage Habit’s attributes has ambiguous. Therefore, the
rule becomes unclear. However, Construct Pythagoras Tree successfully done because
the decision can track based on the trunk of the tree.
Keywords—Pythagoras Tree; ID3; Standard Deviation Reduction; Classification
I. INTRODUCTION
Pythagoras, as we know that it has a right triangle, and the algebraic equation is known as c2=a2+b2. Therefore, Pythagoras has several theorems to prove the equation. However, according to (Teia, 2018) that Pythagoras has a connection to a geometrical context. Pythagoras can assume as a term of the area: The hypotenuse square’s area, which opposite the right angle, is equal to the sum of the square’s area on two legs.
Pythagoras’ geometric when the areas have looping until it can establish the fractal. A fractal tree which is known as Pythagoras Tree. In Data Mining, Pythagoras Tree able to used as a Decision Tree (Ambarsari, Ar Rakhman Awaludin, Suryana, Hartuti, & Rahim, 2019), which alternative visualization for hierarchy (Beck, Burch, Munz, Di Silvestro, & Weiskopf, 2015).
The concept of Pythagoras Tree is the split of data from hypotenuse’s area of a square into the legs of the square area. The purpose of building the Pythagoras Tree is to simplify data association explores.
However, construct data in Pythagoras Tree is need the Decision Tree method, which divides two kinds: Classification and Regression. The Classification method focused on the variables which have category or characters, and Regression for the variables was numeric or continuous. Therefore, in the study about Decision Tree have several methods able to use, such as C4.5 (Handrianto & Farhan, 2019) as a preferred algorithm to Classification, which developed from ID3. Besides, CART, which algorithm uses Classification and Regression at the same time. Also, C5.0 barely used because of its still new algorithm, which expanded of C4.5 (Thariqa, Sitanggang, & Syaufina, 2016).
In the study of (Windia Ambarsari, Avrizal, Doni Sirait, Dwiasnati, & Rahim, 2019) used ID3, which usually used as Classification of Information Gain. The study utilizes Standard Deviation Reduction, which used Coefficient of Variant (CV) to construct Regression for Decision Tree. However, especially for the case, we experiment with used CV for Classification, which the data processing obtained from (Herlinda, 2019).
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10517
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative
Commons Attribution-NonCommercial 4.0 International License. 57
(Herlinda, 2019) to clarify the relation of Instagram to self-esteem towards the user. Result of the study that the influence of Instagram usage on self-esteem is low because it was not able to predict self-esteem level. Processing data on the study used a question and Likert scale to discover relationships. However, the research for Pythagoras Tree applied utilizes a data raw by removing the item obtained by the Likert level. Another element used in the study of (Herlinda, 2019) becomes attributes for Classification Instagram Usage Habit.
II. LITERATURE REVIEW
A. ID3 of Standard Deviation Reduction
As mention before, Standard Deviation Reduction (SDR) used for construct Regression on Decision Tree. Also, SDR based on standard deviation (σ) has beneficial for a numerical dataset on conditional attributes nor target attributes. When σ has zero, instances in a dataset becomes homogeneous. The equation of σ is (Windia Ambarsari et al., 2019):
𝜎 = √1
𝑛∑ (𝑢𝑖 − 𝜇)2𝑛
𝑖=1 (1)
where
𝜇 = 1
𝑛∑ 𝑢𝑖
𝑛𝑖=1 (2)
Meanwhile, the construction of Decision Tree by SDR is by splitting datasets, which purpose to verify the connection between two attributes, namely target and predictor (conditional). Predictor as a determination of Decision Tree establishing.
The standard deviation is calculated recursively for developing each branch, which its subtracted with the previous standard deviation for obtaining SDR. The attribute selected with the most substantial SDR is the best choice in determining the branch. The equation of SDR as follows (Windia Ambarsari et al., 2019):
𝑆𝐷𝑅 = 𝜎 − 𝜎(𝑇, 𝑃) (3)
where
𝜎(𝑇, 𝑃) = (𝑛1
𝑛 𝑥 𝜎(𝑃)1) + (
𝑛2
𝑛 𝑥 𝜎(𝑃)2) (4)
Decision Tree stopped construction when the Coefficient of Variant (CV) is less than the threshold (<0.1).
𝐶𝑉 = 𝜎(𝑃)
𝜇 𝑥 100% (5)
B. Pythagoras Tree
The following theory presented in the article of (Ambarsari et al., 2019), fundamentals of Pythagoras Tree as follow:
1) Pythagoras based on the terms of area, which is the square area of the hypotenuse is equal to the sum of the square area of opposite and adjacent.
2) As mention before, the equation of Pythagoras is c2 = a2 + b2. It is mean that to construct a decision tree based on Pythagoras, which dataset c2 and subset of a2 and b2. Subsets split become legs, based on SDR.
𝐴𝑟𝑒𝑎1 = (𝑎 + 𝑏)2 (6)
𝐴𝑟𝑒𝑎2 = 𝑐2 + 4 (1
2 𝑎𝑏) (7)
𝐴𝑟𝑒𝑎1 = 𝐴𝑟𝑒𝑎2
(𝑎 + 𝑏)2 = 𝑐2 + 4 (1
2 𝑎𝑏)
𝑎2 + 2𝑎𝑏 + 𝑏2 = 𝑐2 + 2𝑎𝑏
𝑎2 + 𝑏2 = 𝑐2 (8)
3) Construct Pythagoras Tree also determined by angle to distinguish opposite and adjacent, which the equation is:
sin 𝜃 = 𝑜𝑝𝑝𝑜𝑠𝑖𝑡𝑒
ℎ𝑦𝑝𝑜𝑡𝑒𝑛𝑢𝑠𝑒 (9)
The Illustration of Pythagoras Tree as follows below:
Fig. 1 Illustration of Pythagoras Tree (da Costa Reis, 2015)
III. PROPOSED METHOD
Classification with regression method for constructing Pythagoras Tree to Instagram Habit Usage datasets have several steps as follows:
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10517
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative
Commons Attribution-NonCommercial 4.0 International License. 58
Fig. 2 Several Steps to Construct Instagram Habit Usage’s Pythagoras Tree
Instagram Habit Usage’s datasets as attributes that have Characters or Categories calculated the Standard Deviation (σ) and the Mean (µ) for the beginning. Calculation Standard Deviation (σ) itself by subtracting each attribute target with a mean (µ), each subtraction of the values given by the power of two.
Split Instagram Habit Usage’s datasets to verify the relation to each attribute of target and predictor. The construct decision tree is doing some training by comparing SDR, whichever is more significant. When the CV > 0.1, the dataset needs subdividing. Splitting datasets based on SDR become leg determination for construct Pythagoras Tree, especially the size of the square area. Also, build branches for construct Pythagoras Tree depends on the angle 𝜃 because it determines which one of two legs is opposite and adjacent.
IV. RESULT AND DISCUSSION
The study attaches 33 datasets, and 7 attributes (Initials, Age, Propose to use Instagram, Download Content, Frequency of using Instagram, The duration of using Instagram, Cumulative of Using Instagram) taken from (Herlinda, 2019), which the result of constructing Pythagoras Tree as in Fig. 3:
Fig. 3 Instagram Habit Usage’s Pythagoras Tree
The first is to determine attribute as the target; the study obtains Age attribute. Observe in Fig. 4, in Instagram Habit Usage’s datasets, there is empty data. The results in empty data not counted. Therefore, the total number of n that should be 33 datasets becomes 32.
Fig. 4 Empty Data In Instagram Habit Usage’s datasets
Calculation of the mean (µ); (20+21+19+19+…+19+18)/32=19.28125, and the
Standard Deviation (σ) = √(0.5166015632 + 2.9541015632 + 0.0791015632 +…+ 1.6416015632) = 1.067543178.
Split the datasets to verify the relation of target and predictor attributes. Conduct training to select attributes that have the highest SDR values. For example, we separate the attribute of Cumulative of Using Instagram, which has SDR = 0.099426768.
Classification
• Determine Standard Deviation (σ) and Mean (µ)
• Determine Standard Deviation for each branch recursively
• Select SDR which has substantial
• SDR stopped to construct when Coefficient of Variant <0.1
Pythagoras Tree
• Standard Deviation for each branch become legs (opposite and adjacent)
• Calculate angle for determine opposite and adjacent to construct Pythagoras Tree
Instagram Habit Usage
• Visualization Data of Decision Tree
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10517
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative
Commons Attribution-NonCommercial 4.0 International License. 59
TABLE 1. CUMULATIVE OF USING INSTAGRAM ATTRIBUTE
Cumulative of Using Instagram
Dataset
s
Standard
Deviation Average Total CV
<0.5 h 0.942809 20 9 4.714045
0.5 to 1
h
0.978019 19 23 5.14747
1 to 2
hrs
2 to 4
hrs
4 to 6
hrs
In Table 1. we split data of <0.5 hours, in which the total sample is 9 instances, and the remaining data is 23—this separate data used as the legs of Pythagoras, namely 9 and 23. As mention before, a total of 32 instances (9+23) means hypotenuse. 23 is adjacent (the value of adjacent is usually higher than the opposite), and 9 is the opposite. Therefore, the calculation angle 𝜃 found by sin 𝜃 =
√9/32=32.02776011. Afterward, do the training again as before, for example, we split data from <0.5 hours.
Based on Tabel 2, Instances that Cumulative of Using Instagram is <0.5 hour, which has divided is:
1) Initials: A1, A2, A11, A17, A21, A24, A28, A30, A31
2) Age: 18 to 21 3) Propose to use Instagram: Communicate
Interactively (CI), Uploading Edited Video (UV), Uploading Content (UC), Surfing (Sf)
4) Download Content: Another person's photo, Video cinematic, Selfie, Objects Photo, Group Photo, Never Download
5) Frequency of using Instagram: 1 to 3 times 6) The duration of using Instagram: <5 minutes, 5 to
15 minutes, 15 to 30 minutes
TABLE 2. SPLIT DATA BASED ON DOWNLOAD CONTENT
Init
ials
Age
Prop
ose
to u
se
Inst
agram
Dow
nlo
ad
Con
ten
t
Freq
uen
cy o
f
usi
ng I
nst
agram
Th
e d
urati
on
of
usi
ng I
nst
agram
Cu
mu
lati
ve o
f
Usi
ng
Inst
agram
A1 20 CI
Another
person's
photo
1 to 3
times
< 5
mins <0.5 h
A2 21 UV
Video
cinematic
1 to 3
times
< 5
mins <0.5 h
A11 21 Sf Selfie
1 to 3
times
< 5
mins <0.5 h
A17 18 CI Selfie
1 to 3
times
15 to
30
mins <0.5 h
A21 20 UC
Objects
Photo
1 to 3
times
< 5
mins <0.5 h
A24 20 Sf
Objects
Photo
1 to 3
times
15 to
30
mins <0.5 h
A28 19 Sf
Group
Photo
1 to 3
times
5 to
15
mins <0.5 h
A30 21 Sf
Never
Download
1 to 3
times
5 to
15
mins <0.5 h
A31 20 CI
Objects
Photo
1 to 3
times
5 to
15
mins <0.5 h
The split data of Table 2 calculated as in Table 1, which SDR= 0.283181521. The result is the split of Download Content, which the classification as one is 1 Another person’s photo, 2 Selfie, 3 Objects Photo, and 1 Group Photo; the other is 1 Video Cinematic and 1 Never Download. The total is 9 instances, which separated as in Table 3.
TABLE 3. DOWNLOAD CONTENT ATTRIBUTE
Download Content
Datasets Standard Deviation Average Total CV
Another
person's
photo
0.880631 19.71429 7 4.466967
Selfie
Selfie
Objects
Photo
Objects
Photo
Group
Photo
Objects
Photo
Video
cinematic 0 21 2 0 Never
Download
Based on Table 3, that CV in Video Cinematic and Never Download has zero. However, it is not <
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10517
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative
Commons Attribution-NonCommercial 4.0 International License. 60
0.1, and the Standard Deviation is 0. It means instances in a dataset becomes homogeneous. The construct Pythagoras become stopped. The illustration of constructing Pythagoras Tree based on data split as follows:
Fig. 5 Data Have Separated in Pythagoras Tree
The result of the construct Pythagoras Tree in Fig. 5 till gains the decision for Instagram Usage Habit as view in Fig. 6:
Fig. 6 Instagram Usage Habit Decision Track
The decision of Pythagoras Tree obtained from the trunk of the tree. The end of the tree's trunk itself (said to be a leaf, but has a more substantial area) is a rule used as a decision tree, in which the mean value is equal to the target value. Therefore, the result for Instagram Usage Habit Decision, which can track based on Pythagoras Tree, is Age 19 years. Tracing is convenient to do when the Pythagorean Tree construction is complete, therefore its capable of forms a rule. Below is the Decision Tree rule for Age 19 years in the Instagram Usage Habit:
Fig. 7 Instagram Usage Habit Rule
Base on Fig. 7 that the Instagram Usage Habit, not getting the right pattern to deciding due to the data obtained, is less. The reason is for ages 19 years that frequency of using Instagram between 1 to 3 times, 4 to 6 times, or 7 to 9 times, for example. Therefore, the rule of the Decision Tree becomes ambiguous. However, in Construction of Pythagoras Tree, successfully done.
V. CONCLUSION AND SUGGESTION
Based on observations from the result and discussion section, SDR is useful for the classification method to constructing Pythagoras Tree. However, the target attribute is must use a numerical variable to gain Standard Deviation (σ) and Meant (µ). Empty data does not affect calculations. Although instances must be discarded, thereby reducing the amount of data.
Although the target for Age 19 years able classified by the Pythagoras Tree. In the study, the Instagram Usage Habit Decision has ambiguous attributes. Therefore, the rule becomes unclear, although attributes are ambiguous, in the development of Pythagoras Tree successfully done because the decision can track based on the trunk of the tree.
Beside discuss of Pythagoras Tree construction, it needs a discussion about Pruning in Pythagoras Tree and how to develop Random Forest Pythagoras Tree for further research.
VI. REFERENCES
Ambarsari, E. W., Ar Rakhman Awaludin, A.,
Suryana, A., Hartuti, P. M., & Rahim, R.
(2019). BASIC CONCEPT PYTHAGORAS
TREE FOR CONSTRUCT DATA
VISUALIZATION ON DECISION TREE
LEARNING. Journal of Applied Engineering
Science, 17(4), 468–472.
https://doi.org/10.5937/jaes17-21960
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10517
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative
Commons Attribution-NonCommercial 4.0 International License. 61
Beck, F., Burch, M., Munz, T., Di Silvestro, L., &
Weiskopf, D. (2015). Generalized pythagoras
trees: A fractal approach to hierarchy
visualization. Communications in Computer
and Information Science.
https://doi.org/10.1007/978-3-319-25117-2_8
da Costa Reis, J. N. (2015). Uma Árvore de Pitágoras
Explorando os Fractais no Ensino Médio.
Ciência e Natura, 37(3), 411–418.
https://doi.org/10.5902/2179460X14636
Handrianto, Y., & Farhan, M. (2019). C.45 Algorithm
for Classification of Causes of Landslides.
SinkrOn, 4(1), 120–127.
https://doi.org/10.33395/sinkron.v4i1.10154
Herlinda, H. (2019). Hubungan Antara Durasi
Penggunaan Instagram Per Hari Dan SELF
ESTEEM Pada Remaja Akhir. Seminar
Nasional Teknologi, 559–564.
Teia, L. (2018). The Pythagorean geometric gear.
Australian Senior Mathematics Journal, 32(1),
54–64.
Thariqa, P., Sitanggang, I. S., & Syaufina, L. (2016).
Comparative analysis of spatial decision tree
algorithms for burned area of peatland in Rokan
Hilir Riau. Telkomnika (Telecommunication
Computing Electronics and Control), 14(2),
684–691.
https://doi.org/10.12928/telkomnika.v14i2.354
0
Windia Ambarsari, E., Avrizal, R., Doni Sirait, E.,
Dwiasnati, S., & Rahim, R. (2019). Regression
Tree Role for Interpret Monetizing of Game
Live Streaming. Journal of Physics:
Conference Series, 1424, 012014.
https://doi.org/10.1088/1742-
6596/1424/1/012014
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10513
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative
Commons Attribution-NonCommercial 4.0 International License. 62
Selection of Outstanding Lecturers with Simple
Additive Weighting Method
Embun Fajar Wati
Universitas Bina Sarana Informatika
West Jakarta, Indonesian
Istikharoh
STIKOM Cipta Karya Informatika
West Jakarta, Indonesian
Tuslaela
STMIK Nusa Mandiri
Central Jakarta, Indonesian
Submitted: Feb 29, 2020
Accepted: Mar 22, 2020
Published: Apr 1, 2020
Abstract— Lecturers who excel have the right to be selected and get promotions and
awards according to their academic performance. However, Sriwijaya State Buddhist
College in Tangerang still uses manual assessment so it requires a lot of time to process
the assessment data. Therefore, we need a method of calculation of SAW (Simple
Additive Weighting) that can be used as a media for performance appraisal of
outstanding lecturers who can facilitate an objective assessment. The SAW calculation
aims to assist in the calculation of several criteria in the assessment of the right achieving
lecturer. This assessment includes performance appraisals that include commitment,
integrity, service orientation, discipline, cooperation, and leadership. In addition to the
criteria, the assessment also includes the Employee Performance Target (SKP)
assessment. The Employee Performance Target is in the form of a total assessment of
lecturer performance. Two criteria, namely performance and SKP will be used as a guide
in calculating the selection of outstanding lecturers. Samples were taken as many as 20
lecturers at Sriwijaya State Buddhist College in Tangerang. The simple additive
weighting method is effectively used in the selection of outstanding lecturers with an
assessment limit of more than 0.88. Of the many candidates, there are three lecturers
with adequate performance and SKP, with grades 0.922, 0.88, 0.94. So that the highest
achieving lecturer with the highest score is 0.94.
Keywords—simple additive weighting; outstanding lecturers; assessment
I. INTRODUCTION
One element in the administration of higher education is lecturers. Lecturers are academic staff who are tasked with implementing Tridharma Higher Education, namely teaching, research and community service. Lecturers are entitled to get promotions and awards according to their work assignments and achievements. Promotions and awards will be obtained based on the performance evaluation of lecturers. Lecturer Performance Assessment is important in managing employee performance. This
is intended to find out how the quality of performance possessed by the lecturers, because lecturers are an important part in the development of an educational institution. Performance appraisal of lecturer staff has several assessment factors namely education, implementation of education, research, implementation of community service, and supporting academic activities of lecturers. By assessing the performance of lecturers, it can be seen that lecturers are high-achieving and qualified. The award will be given to qualified lecturers as a thank you company for the dedication and performance of
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10513
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative
Commons Attribution-NonCommercial 4.0 International License. 63
employees towards the company. And lecturers who are less qualified will be given guidance by educational institutions.
Sriwijaya State Buddhist College in Tangerang is one of the many State Colleges that still uses manual assessment. For lecturers who have the largest number of values, the lecturer is entitled to get an award from the institution. However, the assessment process is still done manually, so it requires a long time to do data processing (Didik Kurniawan, Wamiliana, & Aditya, 2015). The process of assessing lecturers is still influenced by the element of subjectivity from those who choose, so it cannot support the process. It is probable that there was an inaccuracy in evaluating the performance of each lecturer. And the absence of a computerized teacher performance appraisal system so that the assessment of lecturer performance is inaccurate and will affect the results of lecturer selection to be less accurate and precise. From the description above, it shows that the current system does not provide optimal services for higher education management for lecturers and teaching staff (Rajagukguk & Limbong, 2017).
Therefore, we need a Decision Support System (DSS) for the assessment of outstanding lecturers to make it easier to calculate the assessment and can be applied on the web, also can be seen on Android that is already installed in the gadget. Android is a type of Operating System (OS) that makes it easy for customers who want to use (Wati, 2018). The way this decision support system works is almost similar to the way an expert system works. So far, conventional computer devices only function as data processing devices, but with an expert system can produce an information (Wati, Siregar, & Kurniawati, 2018). Similarly, a decision support system that can produce information on the assessment of outstanding lecturers quickly, easily and objectively. The decision support system for the selection of outstanding lecturers uses the Simple Additive Weighting method which is done by creating a paired matrix value for each criterion (Fiqih & Kusnadi, 2017). The drafting of this DSS is expected to solve the problems faced, and produce decision recommendations that can help the Assessment Team to determine who truly deserves the title of lecturer achievement (Mufizar, 2015).
II. LITERATURE REVIEW
A. Relevant Studies
The decision support system for the selection of outstanding lecturers using the Simple Additive Weighting method in the University of Lampung environment has been successfully built to help solve the problem of determining lecturers to make a pretext by carrying out the selection process
objectively based on existing criteria (Didik Kurniawan et al., 2015).
Decision Support System (DSS) for Lecturer Achievement Selection at STMIK Tasikmalaya using the SAW method produces a list of outstanding lecturers with the addition of criteria so as to reduce the level of subjectivity (Mufizar, 2015).
The decision support system for the selection of outstanding lecturers using the SAW Method has been successfully built in the STMIK Budi Darma Medan environment. It can be implemented as an alternative in the objective decision making process (Rajagukguk & Limbong, 2017).
Decision Support System in the selection of outstanding lecturers at the Djadajat Maritime Academy Jakarta using the method namely Simple Additive Weighting (SAW) method can speed up the process of determining the selection of outstanding lecturers with accurate calculations with the results of research that the outstanding lecturers are given to A13 with 14.4 results (Fiqih & Kusnadi, 2017).
B. Decision Support System
According to Holzinger (2011), Decision Support System (DSS) is an intelligent system that includes knowledge-based systems to support decision making activities quickly and accurately (Gustriansyah, 2016). According to Tariq and Rafi (2012), DSS uses data, provides an easy-to-use interface, and allows decision makers to use their own insights (Gustriansyah, 2016). According to Faqih and Irrigation (2014), in other words, a Decision Support System is a computer-based information system that produces various alternative decisions to assist management in dealing with various structured problems using data and models (Windarto, 2017). According to Nofriansyah (2014), Decision support systems (DSS) are usually built to support solutions to a problem or to an opportunity (Badrul, Rusdiansyah, & Budihartanti, 2019).
C. Outstanding Lecturer
In the Law of the Republic of Indonesia Number 14 of 2005 Concerning Teachers and Lecturers, Lecturers with high achievements are lecturers who in the past three years have had useful achievements and can be proud of their origin Universities, and are recognized on a national or international scale (Puspitasari & Ilmi, 2016).
D. Simple Additive Weighting (SAW)
The Simple Additive Weighting (SAW) method is often also known as the weighted sum method. The basic concept of the SAW method is to find a weighted sum of the performance ratings for each alternative on all attributes. According to Jayanti
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10513
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative
Commons Attribution-NonCommercial 4.0 International License. 64
(2015), the SAW Method requires the process of normalizing the decision matrix (X) to a scale that can be compared with all existing alternative ratings (Windarto, 2017).
The formula used to normalize is as follows:
Figure 1. Normalize Formula
Information: Rij = The normalized performance rating from alternative Ai on attribute Cj: i = 1,2, ..., m and j = 1,2, ..., n Max Xij = The biggest value of each criterion i Min Xij = The smallest value of each criterion i Xij = attribute value owned by each criterion Benefit = If the biggest value is the best Cost = If the smallest value is the best
The preference value for each alternative (Vi) is
given the following formula:
Figure 2. Preference Formula
Information: Vi = Ranking for each alternative Wj = Value of ranking weight (of each alternative) rij = Normalized performance rating value A greater value of Vi indicates that the alternative Ai is preferred.
III. PROPOSED METHOD
A. Data Collection
Data collection is done by interviewing the selection of high achieving lecturers and collecting files or documents related to the assessment. After all the data has been collected, the next step is to determine the criteria and minimum value limits for lecturer achievement.
B. Data Analysis
Data analysis using simple additive weighting (SAW) method. SAW by using the formula that has been explained in the literature review chapter. Based on the current document, the assessment of outstanding lecturers uses two criteria, namely performance and the Employee Performance Target (SKP). Each criterion has a weight of 40% and 60%. Based on the specified weight, the assessment of
lecturer achievement has a value limit that is more than 0.88.
IV. RESULT AND DISCUSSION
A. Calculation of SAW
The study was conducted at Sriwijaya State Buddhist High School in Tangerang with stages beginning with data collection by interviewing the assessment team and collecting documents in the form of a list of lecturers' values in which there were various performance criteria and the amount of their grades. The overall assessment is the lecturers who teach at Sriwijaya State Buddhist College. Samples of lecturers taken were 20 people, with the criteria of performance and SKP. Performance includes several sub-criteria, namely commitment, integrity, service orientation, discipline, cooperation, and leadership. Whereas meant by SKP is employee performance target.
Because performance appraisal there are six types of assessment (commitment, integrity, service orientation, discipline, cooperation and leadership). The six types of performance will be calculated on average then will be divided by the maximum value of performance.
The next step is normalization using a predetermined formula. After that, calculate the preferences of each criterion using the preference calculation formula. A lecturer with a preference value of more than 0.88 is an outstanding lecturer. The final results of calculations using the SAW method can be seen in the following figure 3.
Figure 3. Assessment Results Data
Example calculation 3 The sample with the largest value will be explained below.
Table 1. Amount of Values
Lecture C1 C2
C 79,33 84,34
B 76,33 80
P 77,5 88
Table 2. Determination of Average Criteria
0
0,2
0,4
0,6
0,8
1
0,8
0,88
0,922
0,72
0,7
0,78
0,8
0,8
0,7
0,9
0,7
0,5
0,7
0,8
0,9
0,94
0,72
0,78
0,7
0,9
A B C D E F G H I J K L M N O P Q R S T
Total
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10513
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative
Commons Attribution-NonCommercial 4.0 International License. 65
Lect Cm I S D Cp L Amount Rata-
Rata
C 91 71 79 78 79 78 476 79,33
B 70 89 90 78 67 64 458 76,33
P 77 89 76 78 77 68 465 77,5
Information : C1 = Performance C2 = SKP Lect = Lecture Cm = Commitment I = Integrity S = Service Orientation D = Discipline Cp = Cooperation L = Leadership
Table 2 shows the six types of performance will be calculated on average. Calculation of the SAW formula starts with normalization for all criteria.
Performance Normalization
C : 79,33
91= 0,871
B : 76,33
91= 0,838
P : 77,5
91= 0,851
SKP Normalization
C : 84,34
88= 0,958
B : 80
88= 0,909
P : 88
88= 1
Table 3. Normalization Results
Lecture C1 C2
C 0,871 0,958
B 0,838 0,909
P 0,851 1
After the normalization results are obtained, the next
step is to calculate using the preference formula.
Weight of each criterion :
C1 = 0,4 (40%) and C2 = 0,6 (60%)
Performance Preference
C : 0,4 x 0,871 = 0,348 B : 0,4 x 0,838 = 0,335 P : 0,4 x 0,851 = 0,340
SKP Preference
C : 0,6 x 0,958 = 0,574
B : 0,6 x 0,909 = 0,545
P : 0,6 x 1 = 0,6
Table 4. Preference Calculation Results
Lecture C1 C2 Results
C 0,348 0,574 0,922
B 0,335 0,545 0,88
P 0,340 0,6 0,94
The final calculation result is using the preference formula, it can be concluded that the highest value is 0.94 achieved by lecturer P.
B. Interface and Class Diagram Produced
The interface generated from the achievement
lecturer assessment system using the Simple Additive
Weighting (SAW) method can be seen in the
following pictures:
Figure 4. Performance Interface
Figure 5. SKP Interface
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10513
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative
Commons Attribution-NonCommercial 4.0 International License. 66
Figure 6. Assessment Data Interface
Figure 7. Normalization and Preference Results
Interface
Figure 8. Class Diagram
The interface in figure 4-7 illustrates a web-based
application that can be opened in an installed
browser. While the entities and attributes created in
the database can be seen in figure 8.
V. CONCLUSION AND SUGGESTION
Assessment of outstanding lecturers with two
criteria, namely performance and SKP can be used as
a reference because it is objective with the final value
that can be known and given a ranking. Value is also
certain with complete data from these two criteria.
One criterion is that performance can represent many
assessments because there are 6 factors that influence
it, namely commitment, integrity, service orientation,
discipline, cooperation, and leadership. Six factors
are searched for the average value then calculated
using SAW to produce a final grade of 20 lecturers
with the lowest value of 0.7 and the highest value of
0.94.
VI. ACKNOWLEDGMENT
The author would like to thank all those who have helped in the completion of this scientific article, especially family, friends, and also the entire SINKRON crew.
VII. REFERENCES
Badrul, M., Rusdiansyah, & Budihartanti, C. (2019).
Application of Simple Additive Weighting
Method for Determination of Toddler Nutrition
Status. Sinkron, 4(1), 19–24.
Didik Kurniawan, Wamiliana, & Aditya, R. C.
(2015). Sistem Pendukung Keputusan
Pemilihan Dosen Berprestasi Menggunakan
Metode Simple Additive Weighting di
Lingkungan Universitas Lampung. Komputasi,
3(2), 90–98.
Fiqih, M., & Kusnadi, Y. (2017). Sistem Pendukung
Keputusan Pemilihan Dosen Berprestasi
Dengan Metode Simple Additive Weighting.
INFORMATION SYSTEM FOR EDUCATORS
AND PROFESSIONALS, 2(1), 41–50.
Gustriansyah, R. (2016). SISTEM PENDUKUNG
KEPUTUSAN PEMILIHAN DOSEN
BERPRESTASI DENGAN METODE ANP
DAN TOPSIS. In Seminar Nasional Teknologi
Informasi dan Komunikasi (SENTIKA) (pp. 33–
40). Yogyakarta.
Mufizar, T. (2015). Sistem Pendukung Keputusan
Pemilihan Dosen Berprestasi Di STMIK
Tasikmalaya Menggunakan Metode Simple
Additive Weighting (SAW). CSRID, 7(3), 155–
166.
Puspitasari, W. D., & Ilmi, D. K. (2016). SISTEM
PENDUKUNG KEPUTUSAN PEMILIHAN
DOSEN BERPRESTASI MENGGUNAKAN
METODE ANALYTICAL HIERARCHY
PROCESS (AHP). Antivirus, 10(2), 56–68.
Rajagukguk, D. M., & Limbong, R. (2017).
Implementasi Metode Simple Additive
Weighting (SAW) Pada Sistem Pendukung
Keputusan Pemilihan Dosen Berprestasi.
MEANS (Media Informasi Analisa Dan
Sistem), 2(2), 124–133.
Wati, E. F. (2018). APLIKASI SISTEM LAYANAN
PESAN ANTAR MAKANAN BERBASIS
ANDROID PADA KEDAI AYAM REMUK,
TANGERANG. JISAMAR (Journal of
Information System, Applied, Management,
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10513
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative
Commons Attribution-NonCommercial 4.0 International License. 67
Accounting and Research), 2(2), 1–12.
Wati, E. F., Siregar, M. H., & Kurniawati, N. I.
(2018). Expert System Diagnosa Penyakit Paru
pada Anak dengan Metode Forward Chaining.
JISICOM, 2, 10–15.
Windarto, A. P. (2017). IMPLEMENTASI
METODE TOPSIS DAN SAW DALAM
MEMBERIKAN REWARD PELANGGAN.
Kumpulan JurnaL Ilmu Komputer (KLIK),
4(1), 88–101.
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10521
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative
Commons Attribution-NonCommercial 4.0 International License. 68
Expert System Detects Problems of Inkmaker
Machine And Mixer With Forward Chaining
M. Sinta Nurhayati
Bina Sarana Informatika University Jakarta
Jakarta, Indonesia
Rachmat Hidayat
Bina Sarana Informatika University Jakarta
Jakarta, Indonesia
Submitted: Mar 6, 2020
Accepted: Mar 22, 2020
Published: Apr 1, 2020
Abstract— Industrial growth is characterized by the ability to progress in the field of
production both in the type of industry, increasing the quality and volume of production
activities. Machine is one of the factors of production which determines the smoothness of
a production process. In order for the production process to run efficiently, a certain part of
the company that is needed to support the maintenance and repair of the machine is called
Maintenance Engineering. PT. Segwerk Ind is a multinational company engaged in printing
inks for the packaging industry with imported raw materials such as dyes, varnish. The
problem of how to maintain the condition of the machine so that it is always prime when
production activities take place, Trouble machines are indeed one of the factors that can
make a company go bankrupt, such as the production process is hampered, a long machine
down time, so that customers will replace their suppliers because they do not want to risk
business they are stunted. Therefore, for the success factor of a company's production
process, it is required to have appropriate preventive and corrective maintenance activities
as well as the problem of lack of competent human resources, therefore we need tools that
can solve these problems. The results displayed by making expert system software show
that the ink machine problem solving can be known based on the historical machine so that
the downtime faced by PT. Segwerk Ind can be pressed as little as possible against machines
in the ink industry such as ink printing machines.
Keywords— detecting ink engine problems; expert systems; forward chaining
I. INTRODUCTION
Industrial growth is characterized by the
advancement capability in the production field in
both industrial types, improvement of quality and
volume of production activities. The machine is one
of the production factors that determine the
smoothness of a production process. For the
production process to run efficiently, a certain part in
the company that supports to do maintenance and
repair to the machine is Maintenance Engineering.
Maintenance and repair system that already exist in
several companies already based Online (CMMS), so
that the production operator can send a Job Order
Online in case of trouble machines, then maintenance
engineering will immediately take action ranging
from the manufacturing to repairs/maintenance, but
this can only be done in a certain shift condition (such
as morning shift).
According to (Pudji W, 2012) the journal PT.
Philips Indonesia is a company engaged in the field
of lighting (lighting / lamps). This company always
produces maximum production results. In the Lamp
Component Factory section especially in the Stem
Glass department there are 3 types of machines that
operate, including Tubing, Flare and Exhaust Cutting
Machine (ECM) machines. The three types of
machines play an important role in producing lighting
components so that we need the best method to avoid
frequent damage. The maintenance includes
corrective maintenance, namely maintenance
activities after the machine is broken and preventive
maintenance, namely machine maintenance activities
to prevent damage
According to the expert system Journal (Putra
Tanjung, 2017) Inverter welding machine is the
development of the generation of conventional
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10521
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative
Commons Attribution-NonCommercial 4.0 International License. 69
welding machines where conventional welding
machines use transformers / coils with a low working
frequency of 50-60Hz, so to get large currents it is
necessary to enlarge the size of the transformer, so the
larger the size the transformer then requires a large
electrical power as well. In the inverter welding
machine the power is produced through the switching
process of the ferrite core transformer where the
switching frequency is very high between 18khz to
100khz so the size of the transformer is very small
and the power generated is large. an expert system
that can analyze the symptoms of damage into a name
damage decision by using the Forward Chaining
method as a tracking method and the Certainty Factor
method as a method for calculating the value of trust
in symptoms given by a welder. Keywords: Expert
System, Inverter Welding Machine, and Certainty
Factor.
According to (Bangun, irawan hadi;rahman
arif;darmawan, 2014) In this study using the
Reliability Centered Maintenance (RCM) II method
to solve the problem. Blowing machine OM is one of
the important machines in the yarn production
process. Blowing OM machine has the highest
downtime so research will focus on the components
of the Blowing OM machine. The results of data
processing showed that the critical components in the
Blowing OM machine based on the frequency of
engine damage and total downtime were flat belt and
spike lattice components. The results of the
maintenance interval analysis show that the type of
surface damage of uneven flat belt rubber has an
optimal maintenance interval of 510 hours, flat rubber
belt loose 260 hours, broken belt flat 580 hours, wood
spike lattice broken 620 hours, and broken spike
lattice nails 500 hours . From the calculation of the
total optimal maintenance cost, the results show that
the surface damage type of flat rubber flat belt is Rp.
7,973,519.82, loose rubber flat belt Rp.
11,000,673.81, broken flat belt is Rp. 14,061,553.06,
spike lattice wood a fracture of Rp.19,170330.63, and
a broken spike lattice nail of 30,880,512.66.
Reliability Centered Maintenance (RCM) II method
compared to the total previous maintenance costs
there was a decrease in maintenance costs in the
Blowing OM machine by 10.27%.
According to (Ahyari, 2019) the Maintenance
Function is to be able to extend the economic life of
existing production machines and equipment and to
make sure that these production machines and
equipment are always in optimal condition and ready
to use for the implementation of the production
process.
According to the journal (Hidayat, 2017)In
making expert systems, it is necessary to make a
decision on how to make a decision making system
using the Simple Additive Weighting model. such as
research on decision making for scholarship
acceptance.
Researchers conducted a case research at PT.
Segwerk Ind is a multinational company engaged in
printing ink for packaging industry with imported raw
materials such as dyes, varnish, where researchers
found the problem of how to maintain the condition
of the machine to be always prime during production
activities. For the sake of maintaining the smooth
production process that will take place then the
authors build an expert system with forward chaining
method, so that the downtime can be suppressed as
small as possible against the existing machines in the
ink industry such as Mixer machine.
II. LITERATURE REVIEW
A. Expert System
According to (Yulianti, 2016 ) the expert system
is one of the fields of artificial intelligence (AI) that
seeks to adopt human knowledge to computers,
combine knowledge and search data to solve
problems that normally require human expertise.
B. Forward Chaining method
According to (siswanto, 2000) The Forward-
Channing method is sometimes called: data-driven
because the inference engine uses information
specified by the user to move the entire network from
'AND' and 'OR' logic until a terminal is determined as
an object. Forward channing starts from a collection
of facts (data) by looking for rules that match the
existing assumptions / hypotheses leading to
conclusions.
C. Backward Chaining Method
According to the journal (Hidayat, Rachmat;
haryanto ; sapinah, 2019) Backward Chaining
Method The backward chaining method is
backward tracking which starts the reasoning from
the conclusion (goal), by looki ng for a set of
hypotheses towards facts that support a set of
hypotheses.Backward Chaining method is the
opposite of forward chaining where it starts
with a hypothesis (an object) and asks for
information to convince or ignore. Backward
chaining inference engines are often called:
"Object-Driven / Goal-Driven".The inference
engine is part of an expert system that tries to
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10521
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative
Commons Attribution-NonCommercial 4.0 International License. 70
use the information provided to find the
appropriate object.
According to the journal (Akil, 2017)
remembering that, backward-chaining is a form of
thought that is controlled by purpose or
goal. Backward-chaining is useful for answering
specific questions such as "What should I do now?"
and "Where are my keys?". Often, the price of the
backward-chaining method is less than linear search
in KB, because the process only touches on the
relevant facts.
.
D. Unified Modeling Language(UML)
According to (rosa a.s;Shalahuddin, 2018)
defines that, "UML is a visual language for modeling
and communication about a system using diagrams
and supporting textures".
According to (Hidayat, 2018) system that can help
improve customer service and quick responses as
well as documentary evidence stored in a database.
Researchers developed a web-based Coustomer
Relationship Management (CRM) system with the
waterfall model.)
According to (Dharwiyanti, 2003) unified
Modeling Language is a graph-based language for
visualizing, specifying, constructing and
documenting from an Object-Oriented (OO) based
software development system. UML is also referred
to as a standard language for the development of a
software that can convey how to make and shape
models - models but does not convey what and when
the model should be made which is one of the
processes of implementing software development.
Whereas UML itself consists of several diagrams,
namely:
1. Use Case diagram
2. Activity diagram
3. Deployment Diagram
4. Component Diagram
E. Entity Relationship Diagram
According to (Priyadi, 2015) defines that, "Entity
Relationship Diagram relations between entities, can
be done using a database modeling called Entity-
Relationship Diagrams (E-R Diagrams)".
The components of Entity Relationship are:
1. Entity
2. Relations
3. Attributes
4. The connecting line
F. Flowchart
According to (Hidayat, 2014) emphasized that,
Flowchart is a graphical depiction of the steps and
sequence of procedures of a program. Flowcharts
help analysts and programmers to solve problems into
smaller segments and help analyze other alternatives
in operation. Flowcharts usually facilitate the
resolution of a problem, especially problems that
need to be studied and evaluated further. Flowchart is
a form of image / diagram that has one or two
directions flow sequentially. Flowcharts are used to
represent and design programs. Therefore the
flowchart must be able to represent the components
in the programming language)
G. Testing
According to (Ariani.Sukamto, Rosa, 2014)
explained that, "Software testing is an element of a
topic that has a wide scope and is often associated
with verification (verification) and validation
(validation) (V&V)
Testing for validation has the following approaches:
a. White-Box Testing
Test the software in terms of design and program
code whether it is able to produce functions, input and
output in accordance with the requirements
specifications. White box testing is done by checking
the logic of the program code. Making test cases can
follow the testing standards of the programming
standards that should be. Examples of white box
testing are, for example, testing the path (by tracing)
looping in programming logic. Testing of the
documentation made must also be done so that the
documentation made remains consistent with the
software made.
III. PROPOSED METHOD
A. Expert Objects Determination
The Expert System object aims to transfer the
ability (transferring expertise) of an expert or other
source of expertise into a computer and then transfer
it from a computer to an unskilled user (not an
expert). This process includes four activities, namely:
a. knowledge acquitition
b. knowledge representation
c. knowledge inference
d. knowledge transfer
In the expert system analyzing the engine trouble
there are several expert objects that are the source of
the builder of this system, including:
1. Expert I, served as Maintenance
Engineering Coordinator
2. Expert II, served as the Head of the Supply
Chain
3. Expert III, Ir. H. Eddy Suwardi serves as
Director.
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10521
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative
Commons Attribution-NonCommercial 4.0 International License. 71
The results of the Expert interview resulted in a conclusion that engine handling must be carried out as quickly as possible so that downtime can be avoided. The following is a collection of data or facts that are
asked to the user to the final conclusion in the form
of suggestions or solutions from the results of
consultations conducted. The expert table that will be
described will present the relationship between the
machine and the symptoms of the machine and the
level of the problem at hand.
Table 1 Types of Machines
No Machine Code Machine name
1 MX Mixer Machine
2 INK Ink Maker Machine
Table 2 Trouble Machine
No Trouble
Code
Machine Trouble
1. MX201 Does the machine can't
running the mixer
2. MX202 Whether the emergency
sensor is off
3. MX203 Whether the grounding has
been interlock on
4. MX204 Whether the dust collector
was turned up
5. MX205 Is the overload sensor
indicator off
6. MX206 Is sensor oil level indicator
ok
7. MX207 What is sensor limit
movement on position on
8. INK001 Does the ink formula can't
drop out / ink maker not
running
9. INK002 Whether the emergency
sensor is off
10. INK003 Is the pressure switch on
position on
11. INK004 What is the supply rawmat
sensor on
12. INK005 What is a vertical pump
indicator error / off
13. INK006 Is the monitor display
unconnected with the main
display board
Table 3 Relationship between Machine Symptoms
and Level of Problems
B. Explanation of Algorithm Table Flow Existing Data About The Symptoms Of The
Machine First Made A Rule (Rule), So That In
Solving The Problem Easier To Do. Knowledge Base
Or Rule Base That Is Made To Get Knowledge About
The Level Of Problems That Occur With The
Machine Is Written In The Form Of If-Then (IF-
Then).
Rule 1
IF P201 And P203 And P205 And P207 And P209
And P211 And P213 Then AA
IF P101 And P103 And P105 And P107 And P109
And P111 Then BB
Rule 2:
IF P202 Then AG
IF P102 Then BA
Rule 3:
IF P201 And P204 And P210 Then AB
IF P101 And P104 And P109 And P111 Then BB
C. Expert System Algorithm
Here is an algorithm from the system that the
author designed
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10521
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative
Commons Attribution-NonCommercial 4.0 International License. 72
Fig .2 Expert System Algorithm
Explanation of the logic flowchart function as
follows:
1. The first process is to log into the system
2. Then after the login process is successful the user
will start the consultation process, in the
consultation process the system will enter the
symptoms of complaints experienced by the
machine.
3. The system will display the answer at the
consultation in accordance with the symptoms
that occur on the machine.
4. The results displayed by the system are problem
solving that have been used by experts in handling
the problem (records based on historical
machines)
D. Best first search
search rules that work based on a combination of
the two previous methods. The best first search
method is the combination of the depth first search
method and the breadth first search method and the
breadth first search method by taking advantage of
both methods. At each step of the first best search
process, choose nodes by applying adequate heuristic
functions to each node or node that we choose by
using certain rules to produce a successor. In best first
search, the search is allowed to visit the lower node
and if it turns out that the higher level node has worse
heuristic value
The advantage:
1. Requires a relatively small memory, because
only nodes - nodes on the active path only
2. Find a solution without having to test more
nodes
Disadvantage: Allows getting stuck on optima values
IV. RESULT AND DISCUSSION
Based on the problems that exist in PT Siegwerk
Indonesia in detecting engine trouble, the results of
the expert system software that has been made, are
then tested through software testing techniques using
use case diagrams, Activty diagrams, Entity
Relationship Diagrams and white box testing.
The following diagram
a. Use Case Diagrams manage engine symptom
Fig 4 Use Case Diagrams manage engine symptom
data
Tabel 5. Deskripsi Use Case Diagrams manage
engine symptom data
Use Case
Name
manage engine symptoms data
Goal Admin can see, enter, change
delete, search engine symptoms
data
Pre-
Condition
Admin login
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10521
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative
Commons Attribution-NonCommercial 4.0 International License. 73
Post-
Condition
Data on engine symptoms can be
seen, searched, added, modified
and deleted
Failed end
Condition
Data on engine symptoms fail to
be seen, searched, added, changed
and deleted
Actors Admin
Main
Flow/Basic
Path
1. Enter the engine symptoms data
2. Change the engine symptom
data
3. Erase data on engine symptoms
4. Looking for engine symptoms
data
5 View engine symptom data
b. Use Case Diagrams determine the consultation
solution
Fig 5 Use Case Diagrams determine the consultation
solution
Table 6. Deskripsi Use Case Diagrams determine the
consultation solution
Use Case
Name
Determine the consultation
solution
Goal Determine the engine trouble
solution
Pre-
Condition
Admin to login
Post-
Condition
Admin to login
Failed end
Condition
Failed to determine a solution
Actors admin
Main
Flow/Basic
Path
1. Input the symptoms of the
machine
2. Changing the Fact Conditions
Symptoms
3. Input the fact fact condition
c. Entity Relationship Diagram
Fig 7 Entity Relationship Diagram
Table 7. Machine File Specifications
No Elemen
Data
Akronim Tipe
Data
Size Keterangan
1 Symptom
code
kodegejala Text 6 Primary
key
2 Symptom
name
Namagejala Text 100
3 Machine
type
Jenis Text 15
4 Solution Solusi Text 200
d. White box testing
Table 8. white box testing
No. Input
Condition
Expected
Result
Test
Result
Conclusi
on
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10521
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative
Commons Attribution-NonCommercial 4.0 International License. 74
1. Enter
symptoms
Displays
conditions
where the
rules are in
accordance
with the
knowledge
base so
that it
displays
the results
of the
consultatio
n
Showing
that the
provision
s of the
rules
accordin
g to the
knowled
ge base
so that it
displays
the
results of
the
consultat
ion
Path
worked
on
V. CONCLUSION AND SUGGESTION
Conclusion
Based on the problems that researchers found in detecting engine trouble, it can be summarized as follows:
a. How to maintain the condition of the machine so that it is always prime when production activities take place
b. Trouble machining is indeed one of the factors that can make a company go bankrupt, such as the production process is hampered, the machine down time is long, so that customers will change their suppliers because they do not want to risk their business being hampered
Suggestions
Suggestions that researchers are expected to be able to further improve the results of delivery:
a. For the success factor of a company's production process, it is required to have appropriate preventive and corrective maintenance activities as well as the problem of lack of competent human resources, therefore we need tools to solve these problems.
b. Build expert system-based computer software, so the downtime can be pressed as little as possible against the machine
c. It takes time for special training for the production team so that they can use this program and also inform the symptoms of new machines outside of existing ones.
d. Maintenance of software (software) and computer hardware (hardware) is needed so that
the system is protected from damage periodically.
VI. REFERENCES
Ahyari, A. (2019). Manajemen Produksi
Pengendalian Produksi Buku 2 (4th ed.). BPFE
Yogyakarta.
Akil, I. (2017). Analisa Efektifitas Metode Forward
Chaining dan Backward Chaining pada sistem
pakar. 13.
http://ejournal.nusamandiri.ac.id/ejurnal/index.
php/pilar/article/view/333/267
Ariani.Sukamto, Rosa, M. S. (2014). Rekayasa
Perangkat Lunak Terstruktur dan Berorientasi
Objek. Informatika.
bangun, irawan;rahman arif;darmawan, Z. (2014).
Production Machine Maintenance Planning
With Reliability Centered Maintenance (Rcm)
Ii In Blowing Om Machine. Jurnal Rekayasa
Dan Manajemen Sistem Industri, 2, 997.
http://jrmsi.studentjournal.ub.ac.id/index.php/j
rmsi/article/view/145/178
Dharwiyanti, S. (2003). Pengantar Unified Modeling
Language (UML). http://rosni-
gj.staff.gunadarma.ac.id/Downloads/files/1432
1/10.+Unified+Modeling+Language.pdf
Hidayat, Rachmat; haryanto ; sapinah, pipin.
(2019). Tunagrahita Student Learning Expert
System with Backward Chaining Method at
YKDW 01 Tangerang School. Publications &
Informatics Engineering Research, 3, 32.
https://doi.org/https://doi.org/10.33395/sinkro
n.v3i1.10020
Hidayat, R. (2014). Sistem Informasi Ekspedisi
Barang Dengan Metode E-CRM Untuk
Meningkatkan Pelayanan Pelanggan. JURNAL
SISFOTEK GLOBAL, 4, 41–45.
https://stmikglobal.ac.id/journal/index.php/sisf
otek/article/view/50/52
Hidayat, R. (2017). Simple Additive Weighting
Method As A Decision Support System for
Student Achievement Scholarship Recipients.
Jurnal & Penelitian Teknik Informatika, 2, 13.
https://jurnal.polgan.ac.id/index.php/sinkron/ar
ticle/view/59/40
Hidayat, R. (2018). Design of E-RCM Web-Based
Customer Complaints System with Waterfall
Model at PT. Superior Copyright Technology.
Jurnal & Penelitian Teknik Informatika, 2, 112.
https://jurnal.polgan.ac.id/index.php/sinkron/ar
ticle/view/113/75
Priyadi, Y. (2015). Kolaborasi SQL & ERD dalam
Implementasi Database. Liris.
Pudji W, E. I. (2012). Preventive Maintenance PT
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10521
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative
Commons Attribution-NonCommercial 4.0 International License. 75
Philiphs Indonesia.
Putra Tanjung, R. (2017). Expert system to detect
damage inverter weld machine using
certaintyfactor method. Infotek, 2, 96.
https://ejurnal.amikstiekomsu.ac.id/index.php/i
nfotek/article/view/96/88
rosa a.s;Shalahuddin, M. (2018). Rekayasa
Perangkat Lunak Terstruktur dan Berorientasi
Objek (Revisi). http://rosa-
as.id/buku/df.php?df=11
siswanto. (2000). Artificial Intelligence. Budi Luhur.
https://app.box.com/s/y26n4gx6s4fha6oo5mnb
Yulianti, W. (2016). Aptitude Testing Based On
Case-Based Reasoning In Expert System To
Determine Interests And Talent Of Elementary
School Students. Urnal Teknologi Dan Sistem
Informasi UNIVRAB, 1, 104–118.
http://jurnal.univrab.ac.id/index.php/rabit/artic
le/view/28/19
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10526
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative
Commons Attribution-NonCommercial 4.0 International License. 76
The Use of Apriori Algorithm in the Formation
of Association Rule at Lotteria Cibubur Ovi Liansyah
STMIK Nusa Mandiri Jakarta
Jakarta, Jl. Melati 3, RT.002/RW.012, Jatiwaringin, Kec.
Pondokgede, Kota Bks, Jawa Barat
Henny Destiana
UBSI Jakarta
Jakarta, Jl. Kamal Raya No.18, RT.6/RW.3,
Cengkareng, Kota Jakarta Barat
Submitted: Mar 11, 2020
Accepted: Mar 21, 2020
Published: Apr 1, 2020
Abstract— Lotteria as one of the franchises that produce sales data every day, has not been
able to maximize the utilization of that data. The sale data storage is still not optimal. By
utilizing sales transaction data that have been stored in the database, the management
can find out the menus purchased simultaneously, using the association rule. Namely,
data mining techniques to find the association rules of a combination of items. The process
of searching for associations uses the help of apriori algorithms to produce patterns of the
combination of items and rules as important knowledge and information from sales
transaction data. By using the minimum support parameters, the minimum and the month
period of the sales transaction to find the association rules, the data mining application
generates association rules between items in April 2019, where consumers who buy hot / ice
coffee will then buy float together with support of 16% and 100% confidence. Knowing which
menu products or items are the most sold, thus lotteria Cibubur can develop a sales strategy
to sell other types of menu products by examining the advantages of the most sold menu
with other menus and can increase the stock of menu ingredients.
Keywords— Data mining, Menu, Apriori Algorithm, Database, association rule
I. INTRODUCTION
An application is needed to analyze the market basket
of drug sales transaction data using data mining as a
data analysis technique that can help pharmacies
obtain knowledge in the form of sales patterns within
a certain month period. Data mining applications are
built using linear sequential processes with the PHP
programming language and MySQL database. The
algorithm used as the main process of market basket
analysis is apriori algorithm using the parameters of
minimum support, minimum confidence, and the
period of months of sales transactions to find the
association rules.(Irfiani, 2019)
In increasing the company's turnover can be
done using the Data Mining process, one of which is
to use apriori algorithm. With the apriori algorithm,
association rules can be found which can later be used
as a pattern of purchasing goods by consumers, this
study uses a data repository of 209 records consisting
of 23 transactions and 164 attributes. From the results
of this study, the item named CREAM CUPID
HEART COAT HANGER is the product most often
bought by consumers. (Putra et al., 2019)
Sales of OPPO brand mobile electronic
products have not used data mining implementation,
where all goods that are already in stock in the
showroom must be sold, all cannot be purchased at
the center. To avoid the buildup of stock of mobile
phones that are less interested and know what type of
opposition brands with the most sales in the
opposition SDC store requires an appriori algorithm.
This can be known using apriori algorithm which is
part of data mining.(Kanti & Indrajit, 2017)
By using a priori algorithm, it can produce a
combination pattern of 17 (seventeen) rules with a
support value of 70% and the highest confidence
value of the 17 (seventeen) rules of 93% contained in
the Lock Broken → Disscolour rule.(Pahlevi &
Sugandi, 2019) Evaluation and evaluation of lecturers'
performance conducted at STT Harapan is still not optimal so that the level of awareness of each lecturer to improve their performance is also not optimal. This study aims to apply data mining algorithm C4.5 in
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10526
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative
Commons Attribution-NonCommercial 4.0 International License. 77
determining the track record of lecturer performance based on 3 input attributes (publication, service and teaching) and 1 output attribute that has "less" "sufficient" and "high" performance. The results of the study are decision trees and rules that provide information on the results of the evaluation of the performance of STT Harapan Medan lecturers in implementing the tri darma of tertiary institutions.(Rismayanti et al., 2019).
II. LITERATURE REVIEW
Data Mining discuss extracting or gathering useful
information from data sets. Information that is usually
collected is hidden patterns in data, relationships
between data elements, or modeling for the purpose
of forecasting data. (Santoso et al., 2016)
Association Rule Mining is a data mining technique
for finding associative rules between combinations of
items. An example of an associative rule from a
purchase analysis at a supermarket is identifying how
likely a customer is to buy bread together with milk.
(Saw, 2019)
Apriori algorithm is the most famous algorithm for
finding high frequency patterns. Apriori algorithm is
divided into several stages called narration or pass.
Formation of itemset candidates, candidates are
formed from a combination (k1) -itemset obtained
from the previous iteration. One way of apriori
algorithm is to prune candidates whose subsets
containing k-1 items are not included in a high-
frequency pattern with a length of k-1. (Saw, 2019)
This stage is a research sequence design from
beginning to end in conducting research:
1. First study
The initial study of this research is by looking
from studying the problem to be investigated.
Then determine the scope of the problem, the
background of the problem, and study some
literature related to the problem and how to find
solutions to the problem
2. Data collection
The data collection of this study the authors
conducted interviews, observations and
documentation. To find out the information
needed, the writer collected sales data in the April
2019 period in Lotteria.
3. Data Processing with Data mining
At this stage the data processing first identify the
problems often faced by the restaurant then
describe these problems in order to obtain a
solution. The next stage is analyzing the problem
using data mining with apriori algorithm to get the
results as a goal to be achieved then it can be used
by the restaurant as knowledge in increasing
product sales.
4. Analysis of Results
At this stage the authors analyze the results using
the Tanagra 1.4 application and apriori algorithm
in connecting data to be tested.(Yanto & Khoiriah,
2015)
Data Mining has been implemented in various
fields, including business or trade, education and
telecommunications. In business, for example, the
results of implementing data miningvusing the
Apriori algorithm can help business people in
decision making policies to what is related to
inventory. For example the importance of the
inventory system in a Pharmacy and what types of
goods are the top priority that must be in stock to
anticipate vacancy of goods. Because of the lack
of stock of goods can affect customer service and
income Pharmacy. Therefore the availability of
various types of medical devices at the Pharmacy
as one of the suppliers of medical equipment,
absolutely to support the smooth distribution to
consumers, so that the activity customer service
going well.(King et al., 1966)
How to find books that are bought together, can
be used association rule, which is a data mining
technique to find the association rules of a
combination of items. The association search
process uses the help of a priori algorithms to
produce patterns of combination of items and
rules as important knowledge and information
from sales transaction data. The results of this
study are in the form of applications to analyze
spending patterns where the resulting pattern can
be used as recommendations in determining sales
strategies by Gramedia.(Listriani et al., 2018)
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10526
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative
Commons Attribution-NonCommercial 4.0 International License. 78
III. PROPOSED METHOD
Figure 1 research stages
IV. RESULT AND DISCUSSION
The following are the menus that fall into the fast
food category in Lotteria Cibubur namely:
TABLE I MENU TYPE No Menu Table
Menu
1. Float
2. Ice Lemon Tea
3. Buldak Burger
4. Cappucinno
5. Hot/Ice Coffe Latte
6. Ico Coco
7. Chicken Puas A
8. Chicken Puas B
No Menu Table
Menu
9. Hot/Ice Cofffe
10. Pepsi
11. Orange Juice
12. Cupbag Bulgogi
13. Tornado
14. Spicy Chicken
15. Lotteria
Based on data on item sales at the Lotteria Cibubur
Store for 1 month, sales obtained transaction patterns
by analyzing the 3 most menu items sold every day,
can be seen in the table below:
A. Formation of 1 Itemset
With a minimum support value of 15%, the support
value of 1 item is obtained using the following formula:
𝑺𝒖𝒑𝒑𝒐𝒓𝒕(𝐀, 𝐁) =∑ 𝐓𝐫𝐚𝐧𝐬𝐚𝐜𝐭𝐢𝐨𝐧 𝐂𝐨𝐧𝐭𝐚𝐢𝐧𝐬 𝐀
∑ 𝐓𝐫𝐚𝐧𝐬𝐚𝐜𝐭𝐢𝐨𝐧 ∗ 𝟏𝟎𝟎%
The following is the calculation of the formation of
1 itemset:
𝑆(Float) =∑ Float
∑ 30=
∑ 22
∑ 30∗ 100% = 73%
𝑆(Ice Lemon Tea) =∑ Ice Lemon Tea
∑ 30
=∑ 13
∑ 30∗ 100% = 43%
𝑆(Buldak Burger) =∑ Buldak Burger
∑ 30
=∑ 1
∑ 30∗ 100% = 3%
𝑆(Cappucino) =∑ Cappucino
∑ 30=
∑ 13
∑ 30∗ 100%
= 43
𝑆(Hot / Ice Coffe Latte) =∑ Hot / Ice Coffe Latte
∑ 30
=∑ 3
∑ 30∗ 100% = 10%
𝑆(Ice Coco) =∑ Ice Coco
∑ 30=
∑ 2
∑ 30∗ 100% = 6%
𝑆(Chicken Puas A) =∑ Chicken Puas A
∑ 30
=∑ 17
∑ 30∗ 100% = 56%
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10526
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative
Commons Attribution-NonCommercial 4.0 International License. 79
𝑆(Chicken Puas B) =∑ Chicken Puas B
∑ 30
=∑ 3
∑ 30∗ 100% = 10%
𝑆(Hot / Ice Coffe) =∑ Hot / Ice Coffe
∑ 30
=∑ 5
∑ 30∗ 100% = 16%
𝑆(Pepsi) =∑ Pepsi
∑ 30=
∑ 2
∑ 30∗ 100% = 6%
𝑆(Orange Juice) =∑ Orange Juice
∑ 30
=∑ 5
∑ 30∗ 100% = 16%
𝑆(Cupbag Bulgogi) =∑ Cupbag Bulgogi
∑ 30
=∑ 1
∑ 30∗ 100% = 3%
𝑆(Tornado) =∑ Tornado
∑ 30=
∑ 1
∑ 30∗ 100%
= 3%
𝑆(Spicy Chicken) =∑ Spicy Chicken
∑ 30
=∑ 1
∑ 30∗ 100% = 3%
𝑆(Lotteria Chicken) =∑ Lotteria Chicken
∑ 30
=∑ 1
∑ 30∗ 100% = 3%
The support value of the 1 item that has been
described can be seen in the table below:
TABLE 2. Value of Support 1 Itemset
No Menu Table
Itemset Amount Support
1. Float 22/30 73%
2. Ice Lemon Tea 13/30 43%
3. Cappucinno 13/30 43%
4. Chicken Puas A 17/30 56%
5. Hot/Ice Coffe Latte 5/30 16%
6. Ico Coco 5/30 16%
7. Chicken Puas A 17/30 56%
No Menu Table
Itemset Amount Support
8. Chicken Puas B 3/30 10%
9. Hot/Ice Cofffe 5/30 16%
10. Pepsi 2/30 6%
11. Orange Juice 5/30 16%
12. Cupbag Bulgogi 1/30 3%
13. Tornado 1/30 3%
14. Spicy Chicken 1/30 3%
15. Lotteria 1/30 3%
With a minimum support of 15%, then the
combination of 1 itemset that does not meet the minimum support will be removed in the following
table:
TABLE 3. Minimum Support 1 Itemset
No Menu Table
Itemset Amount Support
1. Float 22/30 73%
2. Ice Lemon Tea 13/30 43%
3. Cappucinno 13/30 43%
4. Chicken Puas A 17/30 56%
5. Hot/Ice Coffe Latte 5/30 16%
6. Chicken Puas A 17/30 56%
B. Formation of 2 Itemset
Formation of Combination The support value of 2
items is obtained by the following formula:
𝑆𝑢𝑝𝑝𝑜𝑟𝑡(A, B)∑ Transaction Contains A dan B
∑ Transaction∗ 100%
The following is a calculation of 2 item sets:
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10526
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative
Commons Attribution-NonCommercial 4.0 International License. 80
𝑆(Float, Ice Lemon Tea)
=∑ Float, Ice Lemon Tea
∑ 30
=8
30∗ 100% = 26%
𝑆(Float, Cappucino) =∑ Float, Cappucino
∑ 30
=9
30∗ 100% = 30%
𝑆(Float, Chicken Puas A)
=∑ Float, Chicken Puas A
∑ 30
=13
30∗ 100% = 43%
𝑆(Float, Hot / Ice Coffe)
=∑ Float, Hot / Ice Coffe
∑ 30
=5
30∗ 100% = 16%
𝑆(Float, Orange Juice)
=∑ Float, Orange Juice
∑ 30
=1
30∗ 100% = 3%
𝑆(Ice Lemon Tea, Cappucino)
=∑ Ice Lemon Tea, Cappucino
∑ 30
=3
30∗ 100% = 10%
𝑆(Ice Lemon Tea, Chicken Puas A)
=∑ Ice Lemon Tea, Chicken Puas A
∑ 30
=5
30∗ 100% = 16%
𝑆(Ice Lemon Tea, Hot / Ice Coffe)
=∑ Ice Lemon Tea, Hot / Ice Coffe
∑ 30
=2
30∗ 100% = 6%
𝑆(Ice Lemon Tea, Orange Juice)
=∑ Ice Lemon Tea, Orange Juice
∑ 30
=2
30∗ 100% = 6%
𝑆(Cappucino, Chicken Puas A)
=∑ Cappucino, Chicken Puas A
∑ 30
=6
30∗ 100% = 20%
𝑆(Cappucino, Hot / Ice Coffe)
=∑ Cappucino, Hot / Ice Coffe
∑ 30
=1
30∗ 100% = 3%
𝑆(Cappucino, Orange Juice)
=∑ Cappucino, Orange Juice
∑ 30
=2
30∗ 100% = 6%
𝑆(Chicken Puas A, Hot / Ice Coffe)
=∑ Chicken Puas A, Hot / Ice Coffe
∑ 30
=2
30∗ 100% = 6%
𝑆(Chicken Puas A, Orange Juice)
=∑ Chicken Puas A, Orange Juice
∑ 30
=1
30∗ 100% = 3%
𝑆(Hot / Ice Coffe, Orange Juice)
=∑ Hot / Ice Coffe, Orange Juice
∑ 30
=0
30∗ 100% = 0%
The support value of the 2 items that have been
obtained from the description above can be seen in
the table below:
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10526
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative
Commons Attribution-NonCommercial 4.0 International License. 81
TABLE 4. Value of Support 2 Itemset
No Menu Table
Itemset Amount Support
1. Float, Ice Lemon
Tea
8/30 26%
2. Float, Cappucino 9/30 30%
3. Float, Chicken Puas
A
13/30 43%
4. Float, Hot/Ice Coffe 5/30 16%
5. Float, Orange Juice 1/30 3%
6. Ice Lemon Tea,
Cappucino 3/30 10%
7. Ice Lemon Tea,
Chicken Puas A
5/30 16%
8. Ice Lemon Tea,
Hot/Ice Coffe 2/30 6%
9. Ice Lemon Tea,
Orange Juice 2/30 6%
10. Cappucino,
Chickens Puas A
6/30 20%
11. Cappucino, Hot/Ice
Coffe
1/30 3%
12. Cappucino, Orange
Juice 2/30 6%
13. Chickens Puas A,
Hot/Ice Coffe 2/30 6%
14. Chicken Puas A,
Orange Juice
1/30 3%
15. Hot/Ice Coffe,
Orange Juice
0/30 0%
With a minimum support of 15%, the combination of
2 itemsset that does not meet the minimum support
will be removed in the following table:
TABLE 5. Minimum Support 2 Itemset
No Menu Table
Itemset Amount Support
1. Float, Ice Lemon
Tea
8/30 26%
2. Float, Cappucino 9/30 30%
3. Float, Chicken Puas
A
13/30 43%
4. Float, Hot/Ice Coffe 5/30 16%
5. Cappucino, Chicken
Puas A
6/30 20%
C. Formation of 3 Itemset
Formation Combination The support value of 3 items
is obtained by the following formula:
𝑆𝑢𝑝𝑝𝑜𝑟𝑡(A, B, C)∑ Transaction Cont A, B dan C
∑ Transaction∗ 100%
The following is a calculation of 3 item sets:
𝑆(Float, Ice Lemon Tea, Cappucino )
=∑ Float, Ice Lemon Tea, Cappucino
∑ 30
=2
30∗ 100% = 6%
𝑆(Float, Ice Lemon Tea, Chicken Puas A )
=∑ Float, Ice Lemon Tea, Chicken Puas A
∑ 30
=3
30∗ 100% = 10%
𝑆(Float, Ice Lemon Tea, Hot / Ice Coffe )
=∑ Float, Ice Lemon Tea, Hot / Ice Coffe
∑ 30
=2
30∗ 100% = 6%
𝑆(Cappucino, Chicken Puas A, Hot / Ice Coffe )
=Cappucino, Chicken Puas A, Hot / Ice Coffe
∑ 30
=0
30∗ 100% = 0%
𝑆(Cappucino, Chicken Puas A, Ice Lemon Tea )
=Cappucino, Chicken Puas A, Ice Lemon Tea
∑ 30
=0
30∗ 100% = 0%
The support value of the 3 items obtained can be seen
in the table below:
TABLE 6. Itemset Support Value
No Menu Table
Itemset Amount Support
1. Float, Ice Lemon
Tea, Cappucino
2/30 6%
2.
Float, Ice Lemon
Tea, Chicken Puas
A
3/30 10%
3. Float, Ice Lemon
Tea, Hot/Ice Coffe
2/30 6%
4.
Cappucino, Chicken
Puas A, Hot/Ice
Coffe
0/30 0%
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10526
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative
Commons Attribution-NonCommercial 4.0 International License. 82
No Menu Table
Itemset Amount Support
5.
Cappucino, Chicken
Puas A, Ice Lemon
Tea
0/30 0%
Because of the formation of 3 itemset none meet the
15% support value, then the combination of 2 itemset
that meets the formation of the Association.
After all the high frequency patterns have been found,
then the association rules are found that meet the
minimum requirements for confidence by calculating
the confidence of associative rules A → B. Minimum
Confidence = 80%. Confidence value from the rules
A → B is obtained by the following formula:
𝐶𝑜𝑛𝑓𝑖𝑑𝑒𝑛𝑐𝑒(A, B) =∑ Numb transaction contains A & B
∑ TOTAL Transaction A∗ 100%
The following is a calculation of 2 item sets:
𝑆(Float, Ice Lemon Tea) =∑ Float, Ice Lemon Tea
∑ 22
=8
22∗ 100% = 36%
𝑆(Ice Lemon Tea, Float)
=∑ Ice Lemon Tea, Float
∑ 13
=9
13∗ 100% = 69%
𝑆(Float, Cappucino) =∑ Float, Cappucino
∑ 22
=8
22∗ 100% = 36%
𝑆(Cappucino, Float) =∑ Cappucino, Float
∑ 13
=8
13∗ 100% = 61%
𝑆(Float, Chicken Puas A)
=∑ Float, Chicken Puas A
∑ 22
=13
22∗ 100% = 59%
𝑆(Chicken Puas A, Float)
=∑ Chicken Puas A, Float
∑ 17
=13
17∗ 100% = 76%
𝑆(Float, Hot / Ice Coffe)
=∑ Float, Hot / Ice Coffe
∑ 22
=5
22∗ 100% = 22%
𝑆(Hot / Ice Coffe, Float)
=∑ Hot / Ice Coffe , Float
∑ 5
=5
5∗ 100% = 100%
𝑆(Cappucino, Chicken Puas A)
=∑ Cappucino, Chicken Puas A
∑ 13
=6
13∗ 100% = 46%
𝑆(Chicken Puas A, Cappucino)
=∑ Chicken Puas A, Cappucino
∑ 17
=6
17∗ 100% = 35%
Confidence value of 2 items obtained can be seen in
the table below:
TABLE 7. Minimum Confidence 2 Itemset
No Menu Table
Itemset Amount Confidenc
e
1. Float, Ice Lemon
Tea, Cappucino
2/30 6%
2.
Float, Ice Lemon
Tea, Chicken Puas
A
3/30 10%
3.
If you buy Float,
you will buy Ice
Lemon Tea
8/22 36%
4.
If you buy Ice
Lemon Tea, you
will buy Float
9/13 69%
5.
If You Buy Float,
You Will Buy
Cappuccino
8/22 36%
6.
If you buy
cappuccino Then it
will buy Float
8/13 61%
7.
If You Buy A Float
Then You Will Buy
A Satisfied Chicken
13/22 59%
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10526
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative
Commons Attribution-NonCommercial 4.0 International License. 83
No Menu Table
Itemset Amount Confidenc
e
8.
If you buy
cappuccino Then it
will buy Float
13/17 76%
9.
If You Buy A Float
Then You Will Buy
A Satisfied Chicken
5/22 22%
10.
If You Buy Hot /
Ice Coffe Then
You Will Buy
Float
5/5 100%
11.
If You Buy
Cappuccino Then
You Will Buy A
Satisfied Chicken A
6/13 46%
12.
If You Buy A
Satisfied Chicken
A, You Will Buy
Cappucino
6/17 35%
V. CONCLUSION AND SUGGESTION
A. Conclusions
Based on the discussion that has been done with
the Apriori Algorithm and the testing done with the
Tanagra application, the writer draws several
important conclusions. The conclusions are as
follows:
1. Data Mining by using apriori algorithm can be
implemented to analyze the needs of the company,
by using a product sales transaction database
because it can find the trend of itemset
combination patterns so that it can be used as very
valuable information in making decisions to
prepare stock of any type of product needed later
.
2. Application of apriori algorithm in data techniques
Mining is very efficient and can accelerate the process of forming trends in the combination of
itemset patterns of product sales at Lotteria Store
Cibubur, namely with the highest support and
confidence is the combination of Hot products /
Ice and Float are 16% and 100%. This means that
Hot / Ice Coffe and Float products are the most
popular products by consumers.
B. Suggestion
As for the suggestions submitted so that future
research can be better. The suggestions are as
follows:
1. For future research, the object of research
can be expanded again.
2. The research sample used in future research
must be more.
3. Distributing questionnaires to respondents to
obtain the data needed in research can be
reproduced
4. Most product sales sold at Cibubur Lotteria
Store can be determined using apriori
algorithm, by looking at products that meet
the minimum support and minimum
confidence which products are the most sold,
there are difficulties if the data is processed in
large quantities and will require time long
enough.
5. The application of the Apriori Algorithm is
very practical but it needs to be compared
with other algorithms, to test the extent to
which the Apriori Algorithm is still reliable
for processing and finding patterns of
relationships (associations) between items in
large-scale databases.
6. In using the Apriori Algorithm, an
understanding of Association and Data
Mining rules and how to run Tanagra
software applications is needed. 1.4.
7. The application of the Apriori Algorithm is
very effective but it needs to be compared with
other algorithms, to test the extent of the
Apriori Algorithm it is still reliable to look for
association rules in the data mining process
itself.
REFERENCES
Irfiani, E. (2019). Application of Apriori
Algorithms to Determine Associations in
Outdoor Sports Equipment Stores.
SinkrOn, 3(2), 218.
https://doi.org/10.33395/sinkron.v3i2.100
89
Kanti, S., & Indrajit, R. E. (2017). Implementasi
Data Mining Penjualan Handphone Oppo
Store Sdc Tanggerang Dengan Algoritma
Appriori. Seminar Nasional Sains Dan
Teknologi, November, 1–2.
King, D. G., Young, W. E. V., Clarke, A. J.,
Cain, A. J., & Dimbleby, G. W. (1966).
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10526
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative
Commons Attribution-NonCommercial 4.0 International License. 84
The Lanhill Long Barrow, Wiltshire,
England: An Essay in Reconstruction.
Proceedings of the Prehistoric Society, 32,
73–85.
https://doi.org/10.1017/S0079497X00014
341
Listriani, D., Setyaningrum, A. H., & Eka, F.
(2018). PENERAPAN METODE
ASOSIASI MENGGUNAKAN
ALGORITMA APRIORI PADA
APLIKASI ANALISA POLA BELANJA
KONSUMEN (Studi Kasus Toko Buku
Gramedia Bintaro). Jurnal Teknik
Informatika, 9(2), 120–127.
https://doi.org/10.15408/jti.v9i2.5602
Pahlevi, O., & Sugandi, A. (2019). Penerapan
Algoritma Apriori Dalam Pengendalian
Kualitas Produk. Jurnal & Penelitian
Teknik Informatika, 3.
Putra, J. L., Raharjo, M., Sandi, T. A. A.,
Ridwan, R., & Prasetyo, R. (2019).
Implementasi Algoritma Apriori Terhadap
Data Penjualan Pada Perusahaan Retail.
Jurnal Pilar Nusa Mandiri, 15(1), 85–90.
https://doi.org/10.33480/pilar.v15i1.113
Rismayanti, Damayanti, F., & Khairunnisa.
(2019). Penerapan Data Mining Algoritma
C4 . 5 dalam Menentukan Rekam Jejak
Kinerja Dosen STT Harapan Medan. 3,
99–104.
Santoso, H., Hariyadi, I. P., & Prayitno. (2016).
Data Mining Analisa Pola Pembelian
Produk. Teknik Informatika, 1, 19–24.
http://ojs.amikom.ac.id/index.php/semnast
eknomedia/article/download/1267/1200
Saw, B. M. (2019). Penerapan algoritma apriori
dalam pembentukan association rule di
lotteria cibubur.
Yanto, R., & Khoiriah, R. (2015). Implementasi
Data Mining dengan Metode Algoritma
Apriori dalam Menentukan Pola
Pembelian Obat. Creative Information
Technology Journal, 2(2), 102.
https://doi.org/10.24076/citec.2015v2i2.41
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10531
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative
Commons Attribution-NonCommercial 4.0 International License. 85
Optimization of Sugar Salt Fat in the Human
Body Using Genetic Algorithm
Preddy Marpaung1
STMIK Pelita Nusantara
Medan, Indonesia
Arjon Samuel Sitio2
STMIK Pelita Nusantara
Medan, Indonesia
Anita Sindar3
STMIK Pelita Nusantara
Medan, Indonesia
Submitted: Mar 20, 2020
Accepted: Mar 28, 2020
Published: Apr 1, 2020
Abstract— Salt sugar and fat intake in the human body in daily food patterns. To
avoid non-communicable diseases, it is necessary to optimize food intake levels.
Productive age, in this case between 15-55 years have uncontrolled eating patterns to
choose fast food. If this is not monitored, it will cause various diseases. The importance
of evolutionary algorithms for solving complex problems that are difficult to solve
analytically can use mathematical models. The purpose of this study is to determine the
composition of the ideal intake for productive generations with a minimum cost that
must meet the minimum threshold for each nutritional component. Based on the
mathematical model built, an analysis is carried out to find the optimal solution.
Optimization is an effort or activity to get the best results with the given requirements.
There are 5 main components in the Genetic Algorithm, namely Fitness, Value,
Selection, Crossover, and Mutation. The results showed optimization of the intake of
selected chromosomes as the best results, obtained on chromosome offSpring: 10,
Fitness value: 12737.34.
Keywords— Productive Age; Optimization; Genetic Algorithm; Chromosome
I. INTRODUCTION
Comparison between the population of unproductive age (under 15 years and 65 years and above) with the productive age (between 15 to 64 years) economically shows a different level of welfare. Productive age is considered to be able to meet their own needs so that they are able to manage finances independently, so in terms of regulating everyday life patterns. Awareness of the high cost of health has inspired productive ages to improve healthy lifestyles. Consumption of salt sugar and excess fat is a factor in increasing infectious diseases. The limits of consumption of sugar, salt and fat recommended by the Ministry of Health per person per day are: Sugar no more than 50 grams (4 tablespoons); Salt does not exceed 2000 mg of
sodium / sodium or 5 grams (1 teaspoon), and for fat only 67 grams (5 tablespoons of oil). Combined Sugar-Salt-Fat-Intake (SSF) Population at high risk is if the combined intake of Sugar-Salt-Fat (SSF) exceeds the recommended limit (Sugar> 50 g / day, Salt> 5 g / day, and Fat> 67 g /day).
Optimization is one process to achieve an ideal or optimal result of an effective value to be achieved, many methods for finding an optimal solution include; Genetic Algorithms (GA), Ant Colony Optimization (ACO), Particle Swarm Optimization (PSO) and many other optimization methods. Optimization problems are also not free from computational techniques that require faster processing time or cannot be solved precisely, because it will take a long time to achieve the desired goals. The problem of establishment in finance
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10531
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License. 87
(productive age) encourages fast food habits, at productive age is not a problem if it will issue a high budget for healthy eating patterns. The characteristics of problems that require solving genetic algorithms are multi-objective and multi-criteria. Optimization techniques that mimic the process of biological evolution are genetic algorithms. Evolution is a number of individuals in the population.
From generation to generation, individuals act as parents who reproduce offspring. Problems will become complex if the data is large or involves a large number of entities. The study is entitled Implementation of Genetic Algorithms in Cost Optimization to Meet Nutrition Needs. In this study, food data and its nutritional content and prices will be used for testing. From the data will be formed a population with varying amounts. In complex problems complex mathematical formulations are needed which can be very difficult to build or take a long time. Food intake can be maximized considering that each food beverage packaging includes the composition of the product. Knowledge is needed about how much sugar, salt, and fat are consumed correctly.
II. LITERATURE REVIEW
Based on the Indonesian Ministry of Health, recommendations for sugar consumption per day according to age group are 1-3 years old: 2-5 teaspoons, 4-6 years old: 2.5-6 teaspoons, 7-12 years old: 4-8 spoons tea, More than 13 years and adults: 5-9 teaspoons, Elderly: 4-8 teaspoons. The recommended salt consumption limit by the Ministry of Health of the Republic of Indonesia per person per day is 50 grams of sugar or equivalent to 5-9 teaspoons, 5 grams of salt or equivalent to 1 teaspoon, and 67 grams of fat or equivalent to 3 tablespoons of oil . WHO recommends fat intake no more than 30% of total energy intake per day. This is equivalent to 67 grams of fat per day, if the total energy requirement per day is 2000 calories. Or, the equivalent of 5-6 tablespoons of oil per day. The Ministry of Health has issued recommendations for limiting consumption of sugar, salt and fat.
Optimization is an effort or activity to get the best results with the requirements given. Genetic algorithms have been successfully applied to various combinatorial problems such as production planning and scheduling in the manufacturing industry.
After recombination, resultant chromosomes are passed into the successor population. The processes of selection and recombination are then iterated until a complete successor population is produced. At that point the successor population becomes a new source population (the next generation). The GA is iterated
through a number of generations until appropriate topping criteria are reached. These can include a fixed number of generations having elapsed, observed convergence to a best-fitness solution, or the generation of a solution that fully satisfies a set of constraints. There are several evolutionary schemes that can be used, depending on the extent to which chromosomes from the source population are allowed to pass unchanged into the successor population. These range from complete replacement, where all members of the successor population are generated through selection and recombination to steady state, where the successor population is created by generating one new chromosome at each generation and using it to replace a less-fit member of the source population. This is almost complete replacement except that the best one or two individuals from the source population are preserved in the successor population. This scheme prevents solutions of the highest relative fitness from being lost from the next generation through the nondeterministic selection process.
Genetic algorithm is an alternative problem solving for searching, optimization and machine learning. Genetic Algorithm as a branch of Evolution Algorithm is an adaptive method commonly used to solve a value search in an optimization problem. The process in genetic algorithms begins with initialization, which creates random individuals who have a particular set of genes (chromosomes). Chromosome represents the solution of the problem to be solved. Reproduction is done to produce offspring from individuals in the population. Evaluation is used to calculate the fitness of each chromosome. The probabilistic function is used to select individuals who are kept alive. Better individuals (having greater fitness value) have a greater chance of being selected. Genetic Algorithms will involve several operators, namely
1. Evolutionary operations involving the selection process in it.
2. Genetic Operations that involve crossover and mutation operators
There are 5 main components in the Genetic Algorithm, namely:
a) Coding Techniques, Generating Initial Population
Decoding to encode individual forming genes so that the value does not exceed a predetermined range as well as the value of a variable that will be sought as a solution to the problem. If the value of the variable x encoded range becomes [rb ra], the decimal discrete encoding:
x = rb + ra (ra – rb) (g1x10-1+gx10-2+...+gNx10-N) ....1
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10531
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License. 88
explained: ra : upper limit rb : lower limit g : chromosome
b) Generating Initial Population
Techniques in the generation of this initial population there are several ways, including Random Generators, certain approaches (Inserting Specific Values into Genes), Permutation of Genes.
c) Fitness Value
Fitness value states the value of the objective function. The purpose of genetic algorithms is to maximize the value of fitness. If the required minimum value is the inverse fitness value of the function value itself.
Fitness = C – f (x) atau Fitness = 𝐶
𝑓(𝑥)+ 𝜀… … … … … 2
C is a constant ε is a small number determined to avoid zero division and x is an individual.
d) Selection
Selection is used to get a good prospective parent. Select which individuals will be selected for the process of interbreeding and mutation. In certain generations P individuals and their children through crossing and mutation will produce an optimum location so that premature convergence occurs, to avoid forming Linear Fitness Ranking mechanism.
𝐿𝐹𝑅(𝑖) = 𝑓𝑚𝑎𝑥 − ( 𝑓𝑚𝑎𝑥 − 𝑓𝑚𝑖𝑛 ) (𝑅(𝑖) − 1
𝑁 − 1) . .3
LFR (i) = individual LFR value to i, N = Number of individuals in the population. R (i) = i-individual ranking after being sorted from the largest to the smallest Fitness value. The highest fitness value. Fmin = lowest fitness value
e) Crossovers
Crossovers are operators of genetic algorithms that involve two parents to form new chromosomes. Individuals are chosen randomly to cross between 0.6 to 0.95.
f) Mutation
Child chromosomes are mutated by adding a very small random value (mutation step size), with a low probability.
III. PROPOSED METHOD
Steps in research:
IV.
Fig 1. Steps in Research
Step in research Fig.1 explained starts from collecting data (data intake of foods containing fat salt sugar) then analyzing data and designing a food intake optimization system. Implement optimization using genetic algorithms. From the results of the analysis of genetic algorithm calculations obtained optimal optimization of fat salt sugar intake.
The general structure of Genetic Algorithms, 1) Representation of chromosomes 2) Evaluate by calculating fitness 3) Crossover process to get new individuals 4) A process of mutation that increases population variation. 5) The selection process to form a new population.
Data needed in research:
1. Data on normal food intake from each type of sugar, salt, and fat source.
2. Food intake data categories are more than each type of sugar, salt, and fat source.
3. Production data from each type of sugar, salt, and fat source.
4. Data on sugar, fat salt intake from each food group.
5. Data on sugar, fat salt intake for each age group.
Genetic Algotihm
Chromosome
Representation
Evaluation By
Calculating Fitness
Initial Population
Generation
Selection Process
Crosever
Mutation
Analyze The
Results
Optimization
DATA
COLLECTION
Develop Ideal SSF
Intake
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10531
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License. 89
IV. RESULT AND DISCUSSION
Analyze sugar, salt and fat intake:
1. Sugar Intake
WHO recommends consuming sugar per day less than 10 percent of total energy intake, or approximately 25 grams per day for health purposes.
TABLE I. SUGAR INTAKE BY PRODUCTIVE AGE
Sugar Intake Age (years old)
13-18 19-55 >55
Su1 Sugar 8,7 ± 8,6 10,2 ± 9,0 10,1 ±
8,6
Su2 Brown Sugar 7,4 ± 9,4 7,8 ± 9,6 8,1 ±
10,1
Su3 Jelly 12,1 ± 7,2 11,5 ± 7,5 10,3 ±
7,7
Su4 Candy 8,4 ± 6,7 6,2 ± 6,2 5,7 ± 4,3
Su5 Syrop 17,9 ±
10,8
19,7 ±
12,3
16,1 ±
10,5
Su6 Chocolate 12,2 ±
10,6
10,8 ±
11,9 7,6 ± 5,7
Su7 Jelly gelatin 20,2 ±
21,5
15,1 ±
20,9
15,6 ±
25,3
Su8 Honey 14,4 ±
6,6 14,5 ± 8,8
13,7 ±
8,3
Su9 Sweetener 2,9 ± 4,9 5,1 ± 6,8 3,7 ± 5,5
2. Salt Intake WHO recommends sodium (Na) intake <2 g /day or
equivalent to <5 g salt (NaCl) for adulthood.
TABLE II. SALT INTAKE BY PRODUCTIVE AGE
Foods and Drinks Age (years)
13-18 19-55 >55
S1 Cereals And
Preparations 28.4 17.2 9.2
S2 Tubers and processed 0.6 0.3 0.2
S3 Nuts And
Preparations 0.6 0.7 0.8
S4 Vegetables And
Preparations 0.5 0.7 0.8
S5 Fruit And Processed 0.1 0.2 0.2
S6 Meat And Processed 3.1 2.4 1.4
S7 Offal And Processed 0.1 0.1 0.1
S8 Animal And
Processed 9.2 12 13.5
S9 Egg And Processed 1 0.9 0.7
S10 Milk And Processed 0.2 0.2 0.2
S11 Oil And Processed 0.1 0.1 0.1
S12 Sugar, Syrup And
Confectionary 0.2 0.0 0.0
S13 Spices And
Preparations 53.3 62.2 69.7
S14 Drinks 0 0.9 0.8
S15 Foods 0.2 0.1 0.1
2. Fat Intake
WHO recommends fat intake should not exceed 30 percent of total energy to avoid unhealthy weight
gain. In Permenkes No. 30 of 2013 stated that total fat per day must not exceed 67 grams.
TABLE III. FAT INTAKE BY PROUCTIVE AGE
Code Fat Intake Age (years old)
13-18 19-55 >55
F1 Fat supplement 29.7 21.3 15.2
F2 Bulbs 3.5 1.5 0.7
F3 Nuts 14.4 16.4 21.2
F4 Vegetables 0.3 0.4 0.5
F5 Fruits 43.5 0.4 0.5
F6 Meat 43.5 43.5 53.9
F7 Offal 13.1 13.1 15.5
F8 Animal 9.3 9.3 11.6
F9 Egg 11 11 12.5
F10 Milk 5.1 5.1 5.3
F11 Oils 44.4 44.4 50.5
F12 Syrop, Sugar 0.4 0.4 0.5
F13 Spice 0.7 0.7 0.9
F14 Drink Fat 1.4 1.4 1.5
F15 Composit of food 35.8 35.8 36.6
SSF optimization includes analysis of the influence of genetic parameters on objective functions. The effects studied in this study are grouped into 4 sections, namely:
1. Population size on objective values. 2. Number of generations on objective value. 3. Pc changes on objective values. 4. Pm on objective value 5. Normal SSF on objective values.
TABLE IV. SSF VALUE BY PRODUCTIVE AGE
Age SSF Normal > SSF
1 13-18 65.7 34.3
2 19-55 69.2 30.8
3 >55 79.3 20.7
From the data of SSF intake, 10 types of high-value
foods are taken that affect SSF.
TABLE V. FOOD TYPES TABLE MOST AFFECTS SSF
Code Food Name Age (years)
13-18 19-55 >55
F5 Syrop 19,7 ± 12,3
16,1 ± 10,5
18,5 ± 11,8
F7 Jelly, gelatin
15,1 ± 20,9
15,6 ± 25,3
19,0 ± 23,0
F9 Sweetener 5,1 ± 6,8 3,7 ± 5,5
4,5 ± 6,3
Su1 Sereal 28.4 17.2 9.2
Su6 Meal 3.1 2.4 1.4
Su13 Spices 53.3 62.2 69.7
S6 Chocolate 43.5 43.5 53.9
S10 Milk 5.1 5.1 5.3
S11 Oil 44.4 44.4 50.5
At the initial stage of initialization the parameter is to determine the number of chromosomes in a
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10531
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License. 90
population. The generation size is used to determine how many generation sizes are produced. Crossover Probability and Mutation Probability are used to determine the size of the offspring (child chromosome) produced. It also determines the number of SSF food sources used. The length of each chromosome is a random value in the interval the contents of each gene are random values as much as the amount of food. Following initial initialization parameters anatara lain Total Population = 15, Number of generations = 1, Crossover Probability = 0.5, Mutation Probability = 0.1. Extended intermediate crossovers produce offspring from a combination of two parent values. The mutation process is carried out in each gene with a predetermined mutation rate. The mutation rate value determines the number of genes selected for mutation and produces new offspring. The evaluation process is used to calculate the value of Fitness and determine the best Fitness. The higher the Fitness value the better the chromosomes become a candidate solution. Offspring results generated from the results of the reproduction process are crossever and mutation. The parent chromosome in one solution will be counted, the result:
TABLE VI. CALCULATION OF FITNESS PARENT AND
CHROMOSOM OFFSPRING CALCULATIONS
Kromosom Fitness Kromosom Fitness
Parent 1 54386.34 offSpring 1 36829.62
Parent 2 67460.21 offSpring 2 73828.33
Parent 3 56465.43 offSpring 3 37281.11
Parent 4 25737.27 offSpring 4 32749.23
Parent 5 73638.29 offSpring 5 13748.11
Parent 6 58362.27 offSpring 6 31731.81
Parent 7 58291.22 offSpring 7 42818.13
Parent 8 43233.20 offSpring 8 63713.12
Parent 9 47389.81 offSpring 9 13813.11
Parent 10 89241.32 offSpring 10 12737.34
Parent 11 63759.45 offSpring 11 36717.36
Parent 12 48217.35 offSpring 12 83741.32
Parent 13 64829.42 offSpring 13 23134.56
Parent 14 74925.02 offSpring 14 18429.42
Parent 15 46832.13 offSpring 15 62251.02
The selection process is carried out to obtain the best chromosomes that will be made into the next generation's population. The selection process uses the elitist method, stages:
1. Sort all chromosomes based on the highest to lowest Fitness value.
2. Take the top chromosome as much as the initial population.
TABLE VII. SELECTION RESULT
Chromosomes Fitness
offSpring 10 12737.34
offSpring 5 13748.11
offSpring 9 13813.11
offSpring 14 18429.42
offSpring 13 23134.56
Parent 4 25737.27
offSpring 6 31731.81
offSpring 4 32749.23
offSpring 11 36717.36
offSpring 1 36829.62
offSpring 3 37281.11
offSpring 7 42818.13
Parent 8 43233.20
Parent 15 46832.13
Parent 9 47389.81
Parent 12 48217.35
Parent 1 54386.34
Parent 3 56465.43
Parent 7 58291.22
Parent 6 58362.27
offSpring 15 62251.02
offSpring 8 63713.12
Parent 11 63759.45
Parent 13 64829.42
Parent 2 67460.21
Parent 5 73638.29
offSpring 2 73828.33
Parent 14 74925.02
offSpring 12 83741.32
Parent 10 89241.32
After the selection process, the best chromosome
selection is based on the highest fitness value. The
best chromosome results on offSpring 10 = 2737.34.
Fig. 2 Fitness Calculation Results
0
10000
20000
30000
40000
50000
60000
70000
80000
90000
100000
Fitness Fitness
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10531
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License. 91
Fitness Calculation Results determine the best
chromosome. There are 15 data sources, the results of
calculations, can be seen in Figure 2.
V. CONCLUSION AND SUGGESTION Conclusions from conducting research:
1. The number of population N influences x in generating the initial population. The encoding of the values of individual genes does not exceed the range.
2. Chromosomes or individuals are represented by numbers which are codes for food intake of sugar, salt intake. Fat intake and productive age groups. f (x1, x2, x3) with Range [62.2 0.1], the value of x Chromosome 1 = 29.187, Chromosome 2 = 11.04, Chromosome 3 = 27.50. The crossover method used is the extended intermediate crossover method.
3. The best chromosomes are obtained from the selection results from the highest fitness calculation.
VI. ACKNOWLEDGMENT Thank to STMIK Pelita Nusantara for funding the
internal research program October 2019 - March 2020.
VII. REFERENCES
Badrul, M. Rusdiansyah. Budihartanti, C. (2019).
Application of Simple Additive Weighting
Method for Determination of Toddler
Nutrition Status. SinKron: Journal
Publications & Informatics Engineering
Research. Volume 4, Number 1, pp: 19-24.
Bhuvaneswari, P., & Therese, A. B. (2015). Detection
of Cancer in Lung With K-NN
Classification Using Genetic Algorithm.
Procedia Materials Science, 10(Cnt 2014),
433–440.
Hidayat, K. T. (2018). Genetic Algorithm for
Relational Database Optimization in
Reducing Query Execution Time. Scientific
Journal of Informatics, Vol 5, No 1.
Kristiyanti, D. A. (2019). Optimising the Particle
Swam Optimazion Usage for Predicting
Indonesia Presidential Election Result
Period 2019-2024. SinKron: Journal
Publications & Informatics Engineering
Research, Volume 4, Number 1 pp. 32-38.
Kristiadi, D. Hartanto, R. (2019). Genetic Algorithm
for lecturing schedule optimization. JCCS
(Indonesian Journal of Computing and
Cybernetics, Vol.13, No.1, pp. 83~94
Mahmoodabadi, M.J. Nemati, A. R. (2016). A novel
adaptive genetic algorithm for global
optimization of mathematical test functions
and real-world problems. Engineering
Science and Technology an International
Journal.
Muzid, S. (2014). Population Resizing On Fitness
Improvement Fuzzy Evolutionary
Algorithm (PROFIFEA). Prosiding
SNATIF, 471-478.
Purnomo, A. M. (2019). Algoritma Genetika untuk
Optimasi Komposisi Makanan Bagi
Penderita Hipertensi. Jurnal Teknologi dan
Sistem Komputer, vol. 7, no. 1.
Pranata, N. Uluwiyah, A. Sinaga, A. S. RM. Mockler,
A. Ringrod, K. (2018). Assessing the Impact
of Digital Opportunity on Financial Inclusio.
The Eighth Research Dive on Financial
Inclusion, pp. 19-13.
Rahmi, A. (2017), Hibridisasi Algoritma Genetika
Dengan Variable Neighborhood Search
(VNS) Pada Optimasi Biaya Distribusi,
Jurnal Teknologi Informasi dan Ilmu
Komputer (JTIIK), Vol. 4, No. 2, Juni, hlm.
87-96.
Rahmalia, D. (2019). Fertilizer Production Planning
Optimization Using Particle Swarm
Optimization-Genetic Algorithm. Journal of
Information Systems Engineering and
Business Intelligence, vol. 5, no. 2. Pp. 120-
130.
Rahmi, A., Sarwani, M.Z. & Mahmudy, W.F., 2017.
Genetic Algorithms for Optimization of
Multi-Level Product Distribution. Accepted
to International Journal Of Intelligent
Engineering & Systems.
Sinaga, A. S. RM. Zendrato, R. N. (2019). Optimasi
Penugasan Pegawai Menggunakan Metode
Hungarian. Journal of Innovation
Information Technology and Application
(JINITA), vol 1, no. 1 hal. 16-24.
Sinaga, A. S. RM. (2018). Binary Image Object
Identification Using Fuzzy Logic and
Labeling Breadh-First. Journal of
Telematics and Informatics Vol. 6 No. 4.
Sharma, R. V, Pushpneel. E, (2015). Query
Optimization Concepts in Distributed
Database. International Journal of
engineering Technology Science and
Research 2: 60-65.
Sarwani, M.Z., Rahmi, A. (2017). An Adaptive
Genetic Algorithm for Cost Optimization of
Multi-Stage Supply Chain. Accepted to
Journal Of Telecommunication, Electronic
And Computer Engineering.
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10530
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative
Commons Attribution-NonCommercial 4.0 International License. 92
Decision Support System for Achieving
Scholarship Selection by Using Profile
Matching Method
Rani Irma Handayani1st
STMIK Nusa Mandiri Jakarta
Triningsih2nd
Universitas Bina Sarana Informatika
Melia Putri 3rd
STMIK Nusa Mandiri Jakarta
Submitted: Mar 19, 2020
Accepted: Apr 1, 2020
Published: Apr 1, 2020
Abstract— Learning is one of the obligations of students to do in every school
activity where they study. However, sometimes many students are less able to
digest the subject matter delivered by the teacher. Therefore, the school held a
scholarship program for outstanding students. In order to motivate students to
study harder. Achievement scholarships are given with the aim of motivating
students to study harder. Currently the scholarship is not right on target because
it is still done manually and it is not clear the criteria for a student to get an
achievement scholarship. To conduct an assessment in awarding scholarships to
high achieving students use a decision support system to help solve a problem.
For this reason, to conduct an assessment in the awarding of scholarships, a
decision support system using the Profile Matching method is used. Profile
Matching method is one of the methods used in decision making. In this study,
there are several aspects of the assessment for awarding achievement
scholarships, namely the KKM Aspect, the Attendance Aspect, the Behavior
Aspect, the Craft Aspect or the Discipline, the Neatness Aspect.
Keywords— Profile Matching, Decision Support System, Achievement Scholarships
I. INTRODUCTION
Bina Insan Mandiri Vocational School is one of the
High Vocational Schools that provides achievement
scholarships with the aim of motivating students to study
harder. Scholarships are aids to individuals to continue the
education being pursued (Sari 2018). Not all prospective
scholarship recipients will receive a scholarship, only
candidates who meet the established criteria will receive
the scholarship ( rani irma Handayani 2017)
At present the awarding of scholarships is still
done manually and subjectively to the lengthy process
of determining the scholarship (Oktavia 2018). Due to
limited time and limited ability to see all aspects
accurately often leads to mistakes in decision making
(R. I. Handayani 2015)
Therefore, an assessment is needed for awarding
scholarships to high achieving students by using a
decision support system to help solve a problem
(Saryoko, Aziz, and Nurmalia 2020). Methods that can
be used in assessing scholarships for high achieving
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10530
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative
Commons Attribution-NonCommercial 4.0 International License. 93
students, namely the Profile Matching method. This
method is used to avoid mistakes in decision making (Budi
Sudrajat 2018). This profile matching method can also
facilitate schools in decision-making (Indriyani et al.
2019). Decision making is based on aspects or criteria that
have been determined (Apriana 2016)
II. LITERATURE REVIEW
A. Decision support system
Decision support system (DSS) is an interactive
information system that provides information, modeling,
and manipulating data (Kusrini 2007)
B. Profile Matching
In the profile matching process in outline is a process of
comparing between each criteria for each assessment in a
proposal research proposals submitted so the difference in
scores is known (also called GAP (Gross Across
Product)), the smaller the GAP produced, the greater the
weight value which means it has a greater chance for
eligibility / graduation priority. (Apriana 2018)
The steps in profile matching are:
1. Determine the mapping variables Competency gaps
determine the aspects that will be used in processing
employee grades.
2. Calculating the results of the competency Gap mapping
referred to by the Gap here is the different profiles
between employees and the expected standard profile
or can be shown in the formula below:
Gap = Employee profile – Profile matching..(1).
After obtaining a gap in each employee, each
employee's profile is weighted by benchmarking the gap
value weight table.
Then each aspect is grouped into 2 groups, namely the
Core Factor and Secondary Factor groups. Core Factor
calculations are shown using the formula below:
NCF = Σ Nc ... ... ... ... (2)
Σ Ic
NCF = The average value of the core factor
NC = The number of core factor values
IC = Number of core factor items
Meanwhile, the secondary factor calculation is indicated
by the following formula:
NSF = Σ NS ... ... ... ... (3)
Σ IS
NSF = The average value of the secondary factor
NS = Total number of secondary factor values
IS = Number of secondary factor item
After calculating the Core factors and Secondary
factors, then calculate the total value based on the
percentage of core and secondary that is estimated to
affect the performance of each profile. Examples of
calculations can be seen in the formula below:
(x)% NCF (Core Factor Average Value) + (x)% NSF
(Average Factor Secondary Value) = N (Total of
aspects) ......... (4)
Information :
(x)% = The percent value inputted
Finally the Ranking calculation, the calculation can be
shown by the formula below
Ranking = (x)% N1 + (x)% N2 + (x)% N3 ....... (5)
Information:
N1,N2,N3 : Total aspect value calculated
(x)% : Percent value entered
III. DISCUSSION
In the selection of recipients of Scholarship
Assessment for Student Achievement in West Jakarta
Bina Insan Mandiri Vocational School by using the
profile matching method there are several aspects
assessed and from these aspects there are sub criteria,
namely as follows:
Table 1. Aspects of Providing Scholarships for
Outstanding Students
Scholarship aspects
1. KKM Aspects
a. Mathematics
b. Indonesian
c. English
d. Natural science
e. Social studies
f. Computer Skills and Information
Managemen
2. Aspect of Presence
a. Punctuality in coming to school
b. Punctuality in entering class
c. Punctuality in participating in ceremonial
activities
d. Punctuality in participating in practicum
activities
3. Behavior Aspects
a. Respect, respect the teacher or classmates
and keep the classroom atmosphere in
good condition (not making noise,
chatting, or laughing that has nothing to
do with the lesson)
b. Knock on the door, ask for permission to
enter or exit the class, behave politely or
friendly
c. Does not violate the rules
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10530
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative
Commons Attribution-NonCommercial 4.0 International License. 94
d. Obey all commands of the teacher while in
the school environment
e. Not asking before the teacher asks
4. Crafts / Discipline Aspects
a. Full attention during learning (asking or
answering), carrying textbooks and taking
notes
b. Chase and collect tasks on time
c. The student concerned always complies
with the provisions of school entrance after
school and school hours
5. Neatness aspects
a. The students concerned always dress
neatly and politely
b. Based on the following with the
complete uniform attributes that have
been determined from the school
c. Use black shoes and white socks
Table 2. Value of Aspect Sub Criteria
Sub Criteria Value
1 Very less
2 Less
3 Enough
4 Well
5 Very good
A. Competency Gap Mapping
In calculating the Competency gap mapping, the
gap referred to here is the difference between the
scholarship acceptance profile and the student profile or
can be submitted in the formula below:
:
Gap = Student Profile - Scholarship Acceptance
Profile
Calculation of competency gap mapping is based on
existing aspects. The following is the calculation of the
gap for each aspect:
Table 3. Mapping Competency Gap Aspects of
KKM Scholarship Acceptance
Information :
KKM1 : Mathematics
KKM2 : Indonesian
KKM3 : English
KKM4 : Natural science
KKM5 : Social studies
KKM6 : Computer Skills and Information
Managemen
B. Determination Weight of Gap Value
After obtaining a gap in each student, after the
student profile is given a weight value by benchmarking
the gap value gap table..
Table 4. Weight of Gap Value
No Differe
nce
value
weight
s
Information
1 0 5 There is no difference
(competency as needed)
2 1 4,5 Individual competence is 1
level / level
3 -1 4 Individual competencies
lack 2 levels / levels
4 2 3,5 Individual competence is
excess of 2 levels / levels
5 -2 3 Individual competencies
lack 2 levels / levels
6 3 2,5 Individual competencies are
over 3 levels / level
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10530
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative
Commons Attribution-NonCommercial 4.0 International License. 95
7 -3 2 Individual competencies
lack 3 levels / levels
8 4 1,5 Individual competence is
over 4 levels / level
9 -4 1 Individual competencies
lack 4 levels / levels
Table 5. Determination of the Gap Weighting
Aspects of KKM Aspects for Giving Scholarships
C. Calculation and Classification of Core Factors and
Secondary Factors
After determining the weight of the gap value for the
five aspects. These are the kkm aspect, the attendance
aspect, the behavioral aspect, the craft or discipline aspect,
and the neatness aspect in the same way. Then each aspect
is divided into two factors, namely core factor and
secondary factor.
Table 6. Core Factor and Secondary Factor Values of
KKM
No Student’s
Name CF NF NKKM
1 Afifah Husna 4 4 4
2
Antoni
Gunawan 3,66667 4,33333 3,933334
3 Arifia Nur Jauza 3,66667 4 3,800002
4 Dinar Astriani 4 3,66667 3,866668
5
Nadia Nur
Maidah 4 4 4
6 Nur Elisa Fitria 4,33333 4 4,199998
7
Rizky Ega
Pratama 3,66667 4 3,800002
8
Saprina Putri
Rosita 4 4,33333 4,133332
9
Wahyu Febby
Setiawan 3,66667 4,33333 3,933334
D. Scholarship Score Calculation
The score from this process is the score of the
candidate who was submitted to receive the scholarship.
Determination of the score refers to the results of certain
calculations. The calculation can be shown by the
formula below :
Score(x)%Nkkm+(x)%Nkhd+(x)%Nkel+(x)%Nkk+(x)
%Nkrp
Information :
Nkkm: KKM value
Nkhd: Presence Value
Nkel: Behavior Value
Nkk: Value of Crafts / Discipline
Nkrp: Neat Value
Table 7. Assessment Weight at SMK BINA INSAN
MANDIRI
1. Afifah Husna Score
= (40%*4)+(20%*4,8)+(10%*4)+(20%*4,4)+
(1 0%*3,7)
= 1.6 + 0,96 + 0.4 + 0.88 + 0.37
= 4,21
2. Antoni Gunawan Score
= (40%*3,933334)+(20%*4)+(10%*3)+
(20%*3,7)+(10%*4,7)
= 1,5733336 + 0,8 + 0,3 + 0,74 + 0,47
= 3,883333
No
Assessment criteria
Scholarship SMK
BINA INSAN
MANDIRI
Criteria Weight
1 KKM 40%
2 Presence 20%
3 Behavior 10%
4 Craft / discipline 20%
5 Neatness 10%
Amount 100%
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10530
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative
Commons Attribution-NonCommercial 4.0 International License. 96
3. Score Arifia Nur Jauza
= (40%*3,800002)+(20%*3)+(10%* 4,400002)
+(20%*4,7)+(10%*4)
= 1,5200008 + 0.6 + 0,4400002 + 0.94 + 0,4
= 3,900001
4. Dinar Astriani Score
= (40%*3,86668)+(20%*3)+(10%* 3,999998)
+(20%*3)+(10%*5)
= 1.546672 + 0.6 + 0.3999998 + 0.6 + 0.5
= 3,6466718
5. Nadia Nur Maidah Score
= (40%*4)+(20%*2)+(10%*5)+(20%*4)+
(10%*3,3)
= 1.6 + 0.4 + 0.5 + 0.8 + 0.33
= 3.63
6. Nur Elisa Fitria Score
= (40%* 4,199998)+(20%*4)+(10%*4)+
(20%*4,7)+(10%*4)
= 1.6799992 + 0.8 + 0.4 + 0.94 + 0.4
= 4.219999
7. Rizky Ega Pratama Score
= (40%* 3,800002)+(20%*4)+(10%*4)+
(20%*3,6)+(10%*4,3)
= 1, 5200008 + 0,8 + 0,4 + 0,72 + 0,43
= 3,870001
8. Skor Saprina Putri Rosita
= (40%*4,133332)+(20%*4)+(10%*4,2)+
(20%*3)+(10%*4,1)
= 1,6533328 + 0,8 + 0,42 + 0.6 + 0.41
= 3,883333
9. Skor Wahyu Febby Setiawan
= (40%* 3,933334)+(20%*4)+(10%*3)+
(20%*4)+(10%*4)
= 1,5733336 + 0,8 + 0,3 + 0,8 + 0,4
= 3,873334
Thus the one entitled to receive a scholarship is Nur
Elisa Fitria who received the highest final score of
4.219999 and was ranked 1 (First) out of 9 prospective
recipient students at SMK Bina Insan Mandiri West
Jakarta.
Table 8. Final Results and Scores
No Student’s Name Score Ranking
1 Nur Elisa Fitria 4,219999 1
2 Afifah Husna 4,2 2
3 Arifia Nur Jauza 3,900001 3
4 Antoni Gunawan 3,883334 4
5 Saprina Putri Rosita 3,883333 5
6
Wahyu Febby
Setiawan 3,873334 6
7 Rizky Ega Pratama 3,870001 7
8 Dinar Astriani 3,646672 8
9 Nadia Nur Maidah 3,63 9
IV. CONCLUSION
After conducting research on the granting of
scholarships to outstanding students at SMK BINA
INSAN MANDIRI West Jakarta, a number of
conclusions were obtained, namely:
1. There are 5 aspects of the scholarship grading
assessment used by SMK BINA INSAN
MANDIRI West Jakarta in assessing the selection
of scholarship recipients for outstanding students,
namely: KKM, Attendance, Behavior, Crafts or
Discipline, and Neatness.
2. The scholarship selection system application can
be used as a tool for decision making while still
being based on a decision support system using the
Profile Matching method
REFERENCE
Apriana, Veti. 2016. “PENERAPAN METODE
PROFILE MATCHING UNTUK
MENENTUKAN KELAYAKAN PEMBERIAN
PINJAMAN PADA BANK PERKREDITAN
RAKYAT.” Jurnal Moneter III(2): 144–50.
———. 2018. “Penerapan Metode Profile Matching
Untuk Menentukan Pemberian Reward Terhadap
Pelanggan Pada Bisnis Ritel.” Jurnal Pilar Nusa
Mandiri 14(1): 117–22.
http://ejournal.nusamandiri.ac.id/ejurnal/index.ph
p/pilar/article/download/824/pdf.
Budi Sudrajat. 2018. “Pemilihan Pegawai Berprestasi
Dengan Menggunakan Metode Profile
Matching.” Jurnal sinkron 3: 202–10.
Handayani, rani irma. 2017. “SISTEM PENDUKUNG
KEPUTUSAN PEMILIHAN KARYAWAN
BERPRESTASI DENGAN METODE PROFILE
MATCHING PADA PT. SARANA INTI
PERSADA (SIP).” Jurnal Pilar Nusa Mandiri 13.
Handayani, Rani Irma. 2015. “PEMANFAATAN
APLIKASI EXPERT CHOICE SEBAGAI ALAT
BANTU DALAM PENGAMBILAN
KEPUTUSAN (STUDI KASUS: PT. BIT
TEKNOLOGI NUSANTARA).” Jurnal Pilar
Nusa Mandiri 11(1): 53–59.
http://pilar.nusamandiri.ac.id/index.php/pilar/arti
cle/view/93 (March 9, 2018).
Indriyani, Fintri et al. 2019. “Penerapan Metode Profile
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10530
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative
Commons Attribution-NonCommercial 4.0 International License. 97
Matching Sebagai Pendukung.” 1(2): 2–7.
Kusrini. 2007. Andi Offset Konsep Dan Aplikasi Sistem
Pendukung Keputusan.
Oktavia, Petricia. 2018. “SISTEM PENDUKUNG
KEPUTUSAN SELEKSI PENERIMA
BEASISWA DENGAN METODE WEIGHTED
PRODUCT PADA SMP NEGERI 1 PARUNG
BERBASIS WEB.” 3(2): 80–86.
Sari, Retno. 2018. “Jurnal Evolusi Volume 6 No 2 - 2018.”
Jurnal Evolusi 6(2): 57–65.
Saryoko, Andi, Abdul Aziz, and Lia Nurmalia. 2020.
“SELECTION OF EXTRACURRICULAR
ACTIVITIES IN SMK INSAN AQILAH 4
JAKARTA USING PROFILE MATCHING
METHOD.” 16(1): 39–44.
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10538
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative
Commons Attribution-NonCommercial 4.0 International License. 98
Dog Disease Expert System Using
Certainty Factor Method
Linda Marlinda
STMIK Nusa Mandiri
Jakarta, Indonesia
Widiyawati
STMIK Bani Saleh
Bekasi, Indonesia
Wahyu Indrarti
Universitas Bina Sarana Informatika
Jakarta, Indonesia
Reni Widiastuti
Universitas Bina Sarana Informatika
Jakarta, Indonesia
Submitted: Mar 27, 2020
Accepted: Apr 1, 2020
Published: Apr 2, 2020
Abstract— Many animal keepers at home who do not know the disease contained in
the animal's body. Especially for dogs, the lack of information begins to provide care,
hygiene, vaccinations for the health of pet dogs and sickles that will be caused by dogs
to their owners. Expert systems can provide solutions to the lack of information obtained
by pet dog owners, especially dogs. With this expert system, the owner can know the
dog's disease and realize the right prevention in treatment. In this paper 25, physical
symptoms of the disease are used and found 8 types of common dog diseases. Five
options are given to answer the calculation question using each method: no, quite sure,
sure enough, certain, and certainty sure. Accuracy Analysis of each method is tested by
assessing the results of each analysis method based on user feedback. The results of this
study are the application of an expert system that can diagnose dogs using herbal
medicines from plants. The purpose of this study is to implement the certainty factor
method in the diagnosis system of canine diseases that can provide space in providing
value confidence in knowledge. The conclusion in this study will show some questions
as indicators of the characteristics of canine disease, until the final question. The
conclusion of the study using the certainty factor method will show the characteristics
of the disease in dogs. By being obtained from Rabies (0.9) with easy to get angry refuse
nomal food; Hepatitis (0,9) with swelling of the liver occurs; Distemser (0,8) with
stomach part blister and festering discharge from the eye; canine parvorius (0,8) with
loss of appetite and poop there is blood, herpes virus (0,8) with often roared and
complaints, papilomatosis (0,9) with Smelly dog breath, and leptospirosis (0,7) with
complaints, and Dirofilaria Immitis (0,9) with Unstable body temperature.
Keywords—Expert System, Dog Disease, Certainty Factor (CF)
I. INTRODUCTION
One of the animals that are loved by the public is a
dog. Dogs can also get sick, just like humans. Many
dog owners or even people don't know much about a
disease that can be experienced by a dog. Diseases in
dogs themselves can be contagious and not
contagious (Marselena, Labellapansa, & Syukur,
2018). The risk is that there are some dangerous dog
diseases if humans are affected, for example rabies.
Rabies disease affecting a dog before many did not
know by the owner in advance, it was due to the lack
of knowledge of the dog owner about rabies, the
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10538
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative
Commons Attribution-NonCommercial 4.0 International License. 99
limitations of a veterinarian or also the problem of
limited costs (Kristyanto & Suria, 2018). an owner
can not take care of the dog properly, it will have a
bad impact on the dog. It could make the dog stress,
so that it can cause the dog to get sick (Kristyanto &
Suria, n.d.). Of course there are also many factors that
cause a dog to be affected by a disease, for example
due to weather and temperature factors that are
incompatible with the species of one type of dog, or
due to battery or virus infection. In handling dog
disease, different ways of handling and healing the
disease. For treatments such as rabies, for example,
rabies injections must be performed. Similar to those
about diseases in purebred dogs, which are usually
the most affected by this type of skin disease, the
influencing factors can be due to environmental
conditions and natural habitat, therefore as a dog
owner must have knowledge in caring for dogs so as
not to be mistaken in treat (“The Analysis of
Comparison of Expert System of Diagnosing Dog
Disease by Certainty Factor Method and Dempster-
Shafer Method,” 2013).
In the care of dogs, the owner usually still uses and
relies on the expertise of experts manually, due to the
limitations of the owner's knowledge of diseases that
exist in dogs such as skin diseases. Of course for
treatment and treatment to the vet or expert requires
expensive costs, even more so the existence of a
limited vet (Tse, Bullard, Rusk, Douma, & Plourde,
2019). Therefore the need to create an expert system
to help diagnose diseases in dogs, so that it is easier
to deal with a dog that is affected by the disease so
that care and mild treatment can still be done by the
owner himself (Maniaki & Finch, 2018). an expert
system is expected that the users of this expert system
can find out the symptoms in dogs affected by
rabies(Muttaqin, 2019).
II. LITERATURE REVIEW
The system is a study of a particular domain of knowledge that approaches the ability of humans in a field of expertise. Expert System is a knowledge of human knowledge in the field of expertise which is poured in a computer system in the form of applications that can help humans to solve a problem (Marselena et al., 2018)(Septiani & Kuryanti, 2019).
An expert system is a series built to model the capabilities of a human expert who can help solve problems. Expert systems have now been widely used in various fields of expertise including the field of medicine that can help a doctor to diagnose an illness including pet diseases, namely dogs (Hasibuan, Sunandar, Alas, & Suginam, 2017).
Benefits provided by the Expert System, among others:(Suhery, Midyanti, & Hidayati, 2018)
1. Able to store large amounts of data. 2. Can store data with a fairly long period of
time. 3. Complete the calculation more effectively
and efficiently. 4. Can elaborate on answers to all questions
surrounding a particular area of expertise 5. Can present sumsi, the flow of reasoning
which if necessary is used to arrive at the desired answer.
Dogs are one of the animals that can be made friends. Because these animals can be invited to play, live with humans and be invited to socialize with humans. Dogs are the most popular pets in the world. The dog is a mammal and animal wearer of all what is commonly called an omnivore animal. With their unique abilities, adorable behavior and intelligent intelligence that easily captures what has been trained, this animal can help human activities, dogs in various kinds of environments, namely police dogs that can help police to investigate the location. In the case of accidents, breeding, etc., dogs that can herd dogs herd in a large field and also dogs deliver messages or deliver goods with a single nerve that the dog has memorized the road and the destination house of the package. With all the behavior that can be done by a dog, humans can't have a dog at home not only for one dog but can exceed.(Maniaki & Finch, 2018)
The same thing with humans and other creatures, dogs also have limited body stability if not maintained, eating is not orderly, environmental hygiene is less noticed what happens is that the dog can be exposed to an incurable disease or a disease that causes death. As for some diseases in dogs are as follows:(Kim, Choi, & Park, 2018)
1. Canine Parvovirus (CPV), is an important disease in dogs which is usually referred to as malignant vomiting. This disease is very contagious and is the highest cause of death, especially attacking puppies aged 1-3 months.
2. Distemper, dogs are the most common viral disease in dogs and a few truly isolated dogs are not exposed or infected by this virus. This virus is composed of RNA, helical symmetry, envelopes, the virus is rather unstable and its activity can be damaged by heat, dryness, detergents, fat solvents. Clinical symptoms(“Eka Setyarini,” 2013)
a. Inflammation of the mucus in the broom b. High fever and striking the respiratory tract,
digestion, and nervous system 3. Rabies or mad dog is a zoonotic infectious disease
that can infect humans through the bite of a dog
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10538
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative
Commons Attribution-NonCommercial 4.0 International License. 100
that has rabies. Rabies is a very scary and very malignant disease. Rabies directly attacks the central nervous system (brain)(Kurniati, Mubarok, & Fauziah, 2018)
4. Infectious cannie Hepatitis is a disease caused by type 1 and 2 adeno-virus viruses. Clinical symptoms; (Munirah, Suriawati, & Teresa, 2016) a. At first the dog's appetite dropped so it
looked lethargic and feverish with temperatures above 40 degrees. for about a week, then normal body temperature and rise again. dogs in these conditions like cold places.
b. Her eyes were swollen, continued with red eye membranes, watery discharge in the eyes and nose. in this condition dogs drink water more often because they are thirsty and tend to vomit along with pain in the abdomen and when touched on the swelling in the liver
c. Dogs often cough
5. Tapeworm is a disease that attacks dogs caused by several viruses including echinococcud granulosus, dipylidium caminum, taenia taeniaformis.
Clinical Symptoms;
a. The dog looks listless
b. Decreased appetite
c. Having mild diarrhea and abdominal pain
Dog disease along with some of the causes and clinical symptoms or ways of transmission will be the basic data for the development of an expert system using certainty factors.
III. PROPOSED METHOD
Certainty factor (CF) is a method in the field of expert systems in part a clinical parameter value that is given for the first time by the MYCI holder to show the trustworthiness in a particular field of expertise. This method can be used in the medical field to diagnose a disease (Suhery et al., 2018)(Marlinda, Saputra, & Indrarti, 2019)
This uncertainty can be in the form of probability which depends on the outcome of an event. Uncertain results are caused by 2 factors: the uncertain rules of an uncertain user answer to a question raised by a system. This can be very easily seen in systems for diagnosing diseases where the experts cannot define a relationship between symptoms and their causes with certainty, and patients cannot feel things with
certainty, and in the end result in many possible diagnoses(Sihotang, 2014).
The CF method shows a measure of certainty about a fact or rule. CF is a clinical parameter value given by MYCIN to show the amount of trust. The advantage of the CF method is that it can measure something that is certain or uncertain in making decisions in an expert system of disease diagnosis [9]. The basic formula CF (Hasibuan et al., 2017):
(ℎ, 𝑒) = (ℎ, 𝑒) − (ℎ, 𝑒) (1)
CF (h, e) Certainty Factor in the hypothesis is
influenced by evidence e
MB (h, e) Measure of Belief (level of
confidence), is a measure of the
confidence of the hypothesis h
influenced by evidence (symptoms) e
MD (h, e) Measure of Disbelief (level of
uncertainty), is a measure of distrust
of the hypothesis h influenced by
symptoms e.
H The resulting hypothesis or
conclusion (between 0 and 1)
E Evidence or events or facts
(symptoms)
Strengths and Weaknesses of the Certainty Factor Method.(Setyaputri, Fadlil, & Sunardi, 2018)
a. This method is suitable for use in expert systems that contain uncertainty in the field of medicine to diagnose a disease.
b. In one process the calculation can only process 2 data only so that the accuracy of the data can be maintained.
While certainty factor method deficiencies are:
a. Uncertainty modeling using certainty factor calculation methods is usually debated.
b. For data more than 2 pieces, data processing must be done several times
IV. RESULT AND DISCUSSION
Types of Dog Diseases. This knowledge base can identify dog diseases that contain disease names and disease codes.
Table 1. List of Dog Diseases
Disease The symptoms
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10538
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative
Commons Attribution-NonCommercial 4.0 International License. 101
Rabies (P1) Hiding in a quiet place
running to and fro
easy to get angry
refuse nomal food
Salivate
afraid of water
Hepatitis (P2) swelling of the liver occurs
throws up
high dehydration occurs
Fever
Distemser (P3) throws up
Fever
Cough
stomach part
blister and festering
discharge from the eye
fluid out of the nose
Complaints
loss of appetite
out of breath
canine parvorius
(P4)
throws up
high dehydration occurs
lethargy occurred
loss of appetite
poop there is blood
excrement emits a
characteristic odor
Herpesvirus (P5) throws up
often roared
complaints
the puppy does not want to
breastfeed
Papilomatosis
(P6)
Smelly dog breath
shivering
Leptospirosis
(P7)
throws up
high dehydration occurs
Fever
complaints
Shivering
dirofilaria
immitis (P8)
Cough
complaints
out of breath
loss of appetite
Unstable body temperature
Fig. 1. Expert Tree
Calculation CF
Patients choose the answer to symptoms like this:
1. hiding in a quiet place, with the belief "sure"
2. running to and fro, with the belief "Pretty sure"
3. quick to anger, with the belief "sure"
4. saliva, with the belief "sure"
5. refuse normal food, with the belief "Pretty Sure"
From the case examples, it shows that these symptoms are symptoms of Rabies. The patient's case above, shows that these symptoms are symptoms of Rabies
To get the level of confidence, it is calculated using the certainty factor method as follows:
1. Determine the Value of CF Users
If the value of belief has the following values
analysis Pohon Keputusan
Sistem Informasi
Diagnosa Penyakit Anj ing
G1
G2
G3
G4
G5
G6
67
G8
G9
G10
G11
G12
G13
G14
G15
G16
G17
G18
G19
G20
P1
P2
P3
P4
P5
P6
P7
P8
G21
G22
G23
G24
G25
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10538
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative
Commons Attribution-NonCommercial 4.0 International License. 102
"Sure" = 0.8
"Pretty sure" = 0.6
"Not sure" = 0.4
"Don't know" = 0.2
Then obtained values for CF Users for each symptom are as follows
Symptoms 1 = 0.8
Symptoms 2 = 0.6
Symptoms 3 = 0.8
Symptom 4 = 0.8
Symptom 5 = 0.6
2. Determine Expert Cf Value
Seen from the Expert CF Value table, for the problem in the example case above, the Expert CF value for each symptom is obtained
Symptoms 1 = 0.8
Symptoms 2 = 0.8
Symptoms 3 = 0.9
Symptom 4 = 0.9
Symptom 5 = 0.7
3. calculate the CF Value rule as follows:
CF [H, E] = CF (user) * CF (expert)
CF 1 = 0.8 * 0.8 = 0.64
CF 2 = 0.6 * 0.8 = 0.48
CF 3 = 0.8 * 0.9 = 0.72
CF 4 = 0.8 * 0.9 = 0.72
CF 5 = 0.6 * 0.7 = 0.42
4. Finally, describe the value of CF 1 to CF 5
CF combine 1 = CF 1 + CF 2 * (1 - CF 1)
= 0.64 + 0.48 * (1 - 0.64)
= 0.8128
CF combine 2 = CF combine 1 + CF 3 * (1 - CF combine 1)
= 0.8128 + 0.72 * (1 - 0.8128)
= 0.9475
CF combine 3 = CF combine 2 + CF 4 * (1 - CF combine 2)
= 0.9475 + 0.72 * (1 - 0.9475)
= 0.9853
CF combine 4 = CF combine 3 + CF 5 * (1 - CF combine 3)
= 0.9853 + 0.42 * (1 - 0.9853)
= 0.9914
5. Then get a percentage of confidence in an illness
= CF combine * 100%
= 0.9914 * 100%
= 99%
Table 2 If-then table
No IF Then
1 IF G1 AND G2 AND G3
AND G4 AND
G5 AND G6
P1
2 IF G3 AND G5 AND G8
AND G9 AND
G10
P2
3 IF G3 AND G6 AND
G11 AND G12
P3
4 IF G5 AND G6 AND
G11 AND G12 AND G15
AND G16 AND G17
AND G18
P4
5 IF G6 AND G11 AND
G12 AND G16 AND G17
P5
6 IF G11 AND G12 AND
G16 AND G19 AND 20
P6
7 IF G21 AND G22 AND
G23 AND G24
P7
8 IF G21 AND G22 AND
G26
P8
Table 3. Certainty Factor Value Expert
Disease The symptoms C
F
Rabies (P1) Hiding in a quiet place 0,8
running to and fro 0,8
easy to get angry 0,9
refuse nomal food 0,9
salivate 0,7
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10538
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative
Commons Attribution-NonCommercial 4.0 International License. 103
afraid of water 0,8
Hepatitis (P2) swelling of the liver
occurs 0,9
throws up 0,3
high dehydration occurs 0,4
fever 0,4
Distemser (P3) throws up 0,3
fever 0,4
cough 0,6
stomach part 0,8
blister and festering 0,8
discharge from the eye 0,8
fluid out of the nose 0,2
complaints 0,4
loss of appetite 0,4
out of breath 0,3
canine
parvorius (P4)
throws up 0,4
high dehydration occurs 0,2
lethargy occurred 0,4
loss of appetite 0,8
poop there is blood 0,8
excrement emits a
characteristic odor 0,3
Herpesvirus
(P5)
throws up 0,2
often roared 0,8
complaints 0,8
the puppy does not want
to breastfeed 0,9
Papilomatosis
(P6)
Smelly dog breath 0,9
shivering 0,3
Leptospirosis
(P7)
throws up 0,4
high dehydration occurs 0,4
Fever 0,2
complaints 0,8
Shivering 0,6
dirofilaria
immitis (P8)
Cough 0,2
complaints 0,4
out of breath 0,4
loss of appetite 0,8
Unstable body
temperature 0,9
Display Interface
Display questionnaires for symptoms of canine disease
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10538
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative
Commons Attribution-NonCommercial 4.0 International License. 104
V. CONCLUSION AND SUGGESTION
Disease viruses can attack all creatures created by animal gods including dogs. dogs have the right to life with proper care so they are not easily attacked by diseases. if anjig has been attacked by a disease will endanger himself and humans that are around the dog's environment. Therefore we as humans who have been given the advantage of perfect sense we should pay attention to animals including dogs.
References
Hasibuan, N. A., Sunandar, H., Alas, S., & Suginam,
S. (2017). Sistem Pakar Mendiagnosa Penyakit
Kaki Gajah Menggunakan Metode Certainty
Factor. Jurasik (Jurnal Riset Sistem Informasi
Dan Teknik Informatika), 2(1), 29.
https://doi.org/10.30645/jurasik.v2i1.16
Kim, Y. H., Choi, G. J., & Park, C. (2018). Rate of
left ventricular pressure change by Doppler
echocardiography in dogs with chronic mitral
valve disease at different stages of congestive
heart failure. Veterinary Radiology and
Ultrasound, 59(6), 758–766.
https://doi.org/10.1111/vru.12664
Kristyanto, P. H., & Suria, O. (n.d.). Sistem Pakar
Diagnosa Penyakit Kulit Pada Anjing Ras
Dengan Metode Teorema Bayes Expert System
for Diagnosting Dog Skin Diseases Using
Bayes Method, 11–20.
Kristyanto, P. H., & Suria, O. (2018). Sistem Pakar
Diagnosa Penyakit Kulit Pada Anjing Ras
Dengan Metode Teorema Bayes. JMAI (Jurnal
Multimedia & Artificial Intelligence), 2(2), 11–
20. https://doi.org/10.26486/jmai.v2i2.77
Kurniati, N. I., Mubarok, H., & Fauziah, D. (2018).
Sistem Pakar Untuk Mendiagnosa Penyakit
Hewan Peliharaan Menggunakan Metode
Certainty Factor. Jurnal Teknik Informatika
Dan Sistem Informasi, 4(1).
https://doi.org/10.28932/jutisi.v4i1.708
Maniaki, E., & Finch, N. (2018). Chronic kidney
disease in cats and dogs: Managing proteinuria.
In Practice, 40(7), 266–280.
https://doi.org/10.1136/inp.k3410
Marlinda, L., Saputra, D., & Indrarti, W. (2019).
Expert System Identification Of Learning
Patterns The VARK Method. SinkrOn, 3(2),
205.
https://doi.org/10.33395/sinkron.v3i2.10091
Marselena, S., Labellapansa, A., & Syukur, A.
(2018). Penalaran Berbasis Aturan Untuk
Diagnosa Awal Penyakit Anjing Menggunakan
Teorema Bayes. Jurnal RESTI (Rekayasa
Sistem Dan Teknologi Informasi), 2(2), 530–
535. https://doi.org/10.29207/resti.v2i2.454
Munirah, M. Y., Suriawati, S., & Teresa, P. P. (2016).
Design and Development of Online Dog
Diseases Diagnosing System. International
Journal of Information and Education
Technology, 6(11), 913–916.
https://doi.org/10.7763/ijiet.2016.v6.816
Muttaqin, I. A. (2019). Aplikasi Diagnosa Penyakit
Anjing Berbasis Web Menggunakan, 3(1),
372–379.
Septiani, M., & Kuryanti, S. J. (2019). Sistem Pakar
untuk Mendiagnosa Penyakit Saluran
Pernapasan pada Anak, 2(April 2018), 23–27.
Setyaputri, K. E., Fadlil, A., & Sunardi, S. (2018).
Analisis Metode Certainty Factor pada Sistem
Pakar Diagnosa Penyakit THT. Jurnal Teknik
Elektro, 10(1), 30–35.
https://doi.org/10.15294/jte.v10i1.14031
Sihotang, H. T. (2014). Sistem Pakar Mendiagnosa
Penyakit Kolesterol Pada Remaja Dengan
Metode Certainty Factor (Cf) Berbasis Web.
Jurnal Mantik Penusa, 15(1), 16–23. Retrieved
from http://e-
jurnal.pelitanusantara.ac.id/index.php/mantik/a
rticle/view/161
Suhery, C., Midyanti, D. M., & Hidayati, R. (2018).
Sistem Pakar Diagnosa Penyakit Rabies Pada
Anjing Menggunakan Metode Forward
Chaining. Seminar Nasional Teknologi
Informasi Informasi Dan Multimedia 2018, 19–
24.
Eka Setyarini, The Analysis of Comparison of Expert
System of Diagnosing Dog Disease by
Certainty Factor Method and Dempster-Shafer
Method. (2013). International Journal of
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10538
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative
Commons Attribution-NonCommercial 4.0 International License. 105
Computer Science Issues, 10(1), 576–584.
Tse, C., Bullard, J., Rusk, R., Douma, D., & Plourde,
P. (2019). Surveillance of Echinococcus
tapeworm in coyotes and domestic dogs in
Winnipeg, Manitoba. Canada Communicable
Disease Report, 45(7/8), 171–176.
https://doi.org/10.14745/ccdr.v45i78a01
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10524
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative
Commons Attribution-NonCommercial 4.0 International License. 106
Support Vector Machine Parameter
Optimization to Improve Liver Disease
Estimation with Genetic Algorithm
Hani Harafani
STMIK Nusa Mandiri
Jakarta
Submitted: Mar 9, 2020
Accepted: Apr 2, 2020
Published: Apr 2, 2020
Abstract— Liver disease is an important public health problem. Over the past few
decades, machine learning has developed rapidly and it has been introduced for
application in medical-related. In this study we propose Support Vector Machine
optimization parameter with genetic algorithm to get a higher performance of Root Mean
Square Error value of SVM in order to estimate the liver disorder. The experiment was
carried out in three stages, the first step was to try the three SVM kernels with different
combination of parameters manually, The second step was to try some combination of
range parameters in the genetic algorithm to find the optimal value in the SVM kernel.
The third step is comparing the results of the GA-SVM experiment with other regression
methods. The results prove that GA has an influence on improving the performance of
GA-SVM which has the lowest RMSE value compared to another regression models.
Keywords—Support Vector Machine;Genetic Algorithm;Liver;Parameter;Estimation
I. INTRODUCTION
Liver disease is an important public health problem that mostly manifested as abnormal Liver Function Test (Yao, Li, Guan, Ye, & Chen, 2020). Patient with liver disease have been continuously increasing because of many causes such as excessive alcohol consumption(McDermott & Forsyth, 2016), inhale of harmful gases, intake of contaminated food, pickles, drugs (Venkata Ramana, Babu, & Venkateswarlu, 2011), inherited metabolic liver disease (Kelly, 2019), epidemic, and endemic outbreaks by geographical settings (Beeching & Dassanayake, 2019).
The way to diagnose liver disease is by undergoing liver function test(Yao et al., 2020), and liver disease screening by LFT data is helpful for computer aided diagnosis. Over the past few decades, machine learning has developed rapidly and it has been introduced for application in medical-related fields(Yao et al., 2020), like fertility prediction (Harafani & Maulana, 2019), breast cancer prediction (Purwaningsih, 2019), Liver disease screening (Yao
et al., 2020), liver disease diagnosis (Venkata Ramana et al., 2011) and so on. Therefore, in the regression tasks several Machine learning methods are widely applied such as K-NN (Goyal, Chandra, & Singh, 2014),
Linear Regression (Lira, Da Silva, Alves, & Veras, 2014), and Support Vector Machine (Harafani & Wahono, 2015).
Support Vector Machine has advantages in overcoming classification (Support Vector Classifier Machine) and regression (Support Vector Regression Machine) tasks both with linear kernel or nonlinear kernel (Maimon & Rokach, 2010), besides SVM can provide alternative to computational cost (Zaghloul, Hamza, Iorhemen, & Tay, 2020), and good for overcoming the curse of dimensionality (Wang, Wen, Zhang, & Wang, 2014).
However the forecast accuracy is highly affected by three parameter of SVR (Kavousi-Fard, Samet, & Marzbani, 2014) containing 𝐶 (a parameter to tradeoff between training error and regression function flatness), kernel function parameter, and ε
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10524
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative
Commons Attribution-NonCommercial 4.0 International License. 107
(constant value to determine the width of the loss function in SVR), so that there are so many research studies to solve that major issues by suggesting different approaches (Kavousi-Fard et al., 2014) such as Particle Swarm Optimization (Liang, Zou, Li, Junaid, & Lu, 2019), and Genetic Algorithm (Chen, Liang, Hong, & Gu, 2015). In this study genetic algorithm proposed to optimize support vector machine parameters. The objective of this research is to get a higher estimation performance of the liver disease estimation.
II. LITERATURE REVIEW
Support Vector Machine can be imagined as a surface that creates a boundary between points of data plotted in multidimensional that represent examples and their feature value (Lantz, 2015). The goal is to create a flat boundary called a hyperplane which divides the space to create fairly homogenous partition on either side. Support Vector Regression (SVR) is gaining popularity in regression and classification due to its excellent generalization performance (Liang et al., 2019). in regression applications, to extend to nonlinear regression, the SVR kernel function has been used to project the input space into the feature space producing the linear or nearly linear regression hyper surface in the feature space, therefore, the selection of the SVR penalty parameter 𝐶, and the kernel function parameter γ has an important influence on the SVR regression performance.
The basic idea of SVR is to map data from the input space into high dimensional feature with non-linier mapping and to do linear regression into that
space. 𝑥𝑖+1 = 𝑓(𝑥𝑡, 𝑥𝑡−1, 𝑥𝑡−2, … . 𝑥𝑡−(𝑚−1)) where
𝑥𝑖+1 is the predicted value and 𝑥𝑡 is the predicted value. A set of training data for SVM Regression represented as the function below
𝐷 = {(𝒙1, 𝑦1), (𝒙2, 𝑦2), … (𝒙𝑚, 𝑦𝑚)} (1)
Where 𝒙𝑖 is the n-dimensional vector, and y is the real
amount for each 𝒙𝑖.
The standard of Support vector regression is to
use the loss function 𝐿𝜀(𝑦, 𝑓(𝑥)) that describes the
deviation of estimation function from the original
data. Some type of the loss function that can be
extracted from the literature are: linear, quadratic,
exponential, loss function Huber, etc (Raghavendra.
N & Deka, 2014). In this context the loss function 𝜀
insensitive can define as:
𝐿𝜀(𝑦, 𝑓(𝑥)) =
{0 𝑓𝑜𝑟 |𝑦 − 𝑓(𝑥)| ≤ 𝜀
|𝑦 − 𝑓(𝑥)| − 𝜀 𝑎𝑛𝑑 𝑣𝑖𝑐𝑒 𝑣𝑒𝑟𝑠𝑎(2) (2.2)
By using loss function 𝜀 insensitive, first of all
we can found 𝐹(𝒙𝑖) function that can approximate y
vector as actual output and it has the superior error
tolerance from 𝑦𝑖 target to all data training, SVR
makes the mapping of the 𝒙𝑖 input vector to 𝑦𝑖 target
with this regression function:
𝐹(𝑥) = 𝑤 ∗ 𝜙(𝑥) +𝑏 (2.3)
Where w is weighting vector and b ia a bias.
The goal is to estimate w and b parameter from the
function to give the best result according to the data.
Based on the lowest value of w, with minimizing
||𝑤||2 it can maximize the margin, so that the flatness
of the curve along with the complexity of the model
can be ascertained. So the regression problem can be
stated like the following convex optimization
problem:
Minimum function:
𝐿(𝑤, ξ) =1
2||𝑤||2 + 𝑐 ∑ (ξ2𝑖 , ξ′
2𝑖), 𝑐 >𝑖
0
(2.4)
Subject to:
𝑦𝑖 − 𝑤 ∗ 𝜙(𝒙𝑖) − 𝑏 ≤ 𝜀 + ξ𝑖
𝑤 ∗ 𝝓(𝒙𝑖) + 𝑏 − 𝑦𝑖 ≤ 𝜀 + ξ′
𝑖
(2.5)
Ξ𝑖, ξ′𝑖 ≥ 0 (3)
(2.7)
Where ξ𝑖 , dan ξ′𝑖 are the slack variable
introduced to evaluate the deviation of the training
samples outside the 𝜀 insensitive zone or the distance
of the training dataset point from the area where the
error value is less than 𝜀 value will be ignored.
Trade-off between the flatness F(x) with the
quantity of deviation value until greater than 𝜀 value
can be tolerated by oleh C > 0. C is a positif constant
that influences sanction for losses when training
errors occurs.
To solve optimization problem ini minimum
function we can use the Lagrange function of the
objective function with introduce one double set of
the 𝛼𝑖, 𝑑𝑎𝑛 á′𝑖 variable for appropriate constraints.
The optimal condition are exploited at the saddle
point of Lagrange function leading to the
formulation: 𝑚𝑎𝑥 𝛼𝑖𝛼′
𝑖 −
1
2 ∑ 𝛼𝑖
𝑛𝑖,𝑗=1 -𝛼′
𝑖(𝛼𝑗 − 𝛼′𝑗)(𝜙(𝑥𝑖). 𝜙(𝑥𝑗))
−𝜀 ∑ (𝛼𝑖 + 𝛼′𝑖) + ∑ 𝑦𝑖(𝛼𝑖 − 𝛼′𝑖)
𝑛𝑖=1
𝑛𝑖=1 (2.8)
subject to : ∑ (𝛼𝑖 − 𝛼′𝑖) = 0𝑛
𝑖=1
0 ≤ 𝛼 ≤ 𝐶, 𝑖 = 1 … . 𝑛
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10524
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative
Commons Attribution-NonCommercial 4.0 International License. 108
0 ≤ 𝛼′𝑖 ≤ 𝐶, 𝑖 = 1 … . 𝑛 (4)
Some popular kernel function (Liu, Tian,
Chen, & Li, 2013) are used to solve nonlinear
mapping problem in SVR namely :
1. Polynomial : 𝐾(𝑥𝑖, 𝑥𝑗) = (𝑥. 𝑦 +
1)𝑑 (2.10)
2. Linear : 𝐾(𝑥𝑖, 𝑥𝑗) =
𝑥𝑖𝑇. 𝑥𝑗
(2.11)
3. Sigmoid : 𝐾(𝑥𝑖, 𝑥𝑗) = tanh(𝛽𝑥𝑇𝑥𝑖 +
𝛽1) (2.12)
4. Radial Basis Function (RBF) : 𝐾(𝑥𝑖, 𝑥𝑗) =
exp(−𝛾‖𝑥𝑖 − 𝑥𝑗‖2
)(2.13)
In this study RBF kernel is used as the most
widely used kernel by the world researcher, thus 𝐹(𝑥)
function of SVR becomes into the following function:
𝐹(𝑥) = ∑ (𝛼𝑖 − 𝛼′𝑖)𝐾(𝑥𝑖𝑥𝑗) + 𝑏𝑖 (5)
Genetic algorithm according to Holland in (Zhang &
Wang, 2018) is a global heuristic search technique
that attempts to emulate the mechanic of natural
evolution and the principles of survival of the fittest,
and also, according to Goldberg in (Bhuvaneswari &
Therese, 2015), Genetic algorithm belong to the
larger class of evolutionary algorithm, which
generate solutions to optimization problems using
techniques inspiring by natural evolution, such as
inheritance, mutation, selection, and crossover. GA
contains a population of individuals, each of which
has a known level of fitness (Raikar, Wang, Shih, &
Hong, 2016). The population is evolved through
successive generations; the individuals in each new
generations are bred from the fitter individuals of the
previous generation. The process continues through successive generations until the satisfactory
conditions.
III. PROPOSED METHOD
In this study, we use secondary data from UCI Machine Learning Repository namely Liver Disorders Dataset from BUPA medical research. The dataset consists of 345 rows and 7 columns. Each row corresponds to one human male subject (McDermott & Forsyth, 2016). The first 5 columns are integer-valued and represent the results of various blood test which may be of use in diagnosing alcohol related liver disorders. The 6th columns is a real-valued and represents the number of alcoholic drinks taken per day by the subject, by self reported. The last column is the “selectors” that split the dataset into training and testing subsets. It was created by the BMRDL
researchers. The attributes of the dataset namely MCV (Mean Corpuscular Volume), ALKPHOSE (Alkaline Phospotase), SGPT (Alanine Aminotransferase), SGOT (Aspartate Aminotransferase), GAMMAGT (Gamma Glutamyl Tranpepsidase), drinks number of half-pint equivalent of alcoholic beverages drunk per day, and the selector to split the dataset into training and testing subsets. However, liver disorders datasets has been misinterpreted by many studies for classification tasks for which the classification target is the last attribute of the datasets. Therefore (McDermott & Forsyth, 2016) suggested to use the 6th attribute as a target of regression tasks.
In the first step we delete the last attribute of the dataset since we use 10-fold cross validation method to split the dataset into 90% of training and 10% of testing. Then the data is trained and tested 10 times by the SVM with manually optimize parameter of the SVM kernels.
The Second step, the parameters of SVM kernel will optimized directly with genetic algorithm 10 times. So the SVM RMSE value can be compared and the means of all iteration between SVM and SVM-GA can be compared with t-test. After that the last step is to compare the lowest value of RMSE SVM with another regression method.
Fig1. Proposed Method
IV. RESULT AND DISCUSSION
In the first experiment the parameters of the SVM kernels were optimized manually. The kernel consist of (dot, polynomial, and RBF). In the dot kernel C and ε parameter are tried 10 times manually, and then we assign 10 minimum and maximum range value to the genetic algorithm to find the optimum parameters values to get the lowest RMSE. In the manual
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10524
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative
Commons Attribution-NonCommercial 4.0 International License. 109
experiment, we got the lowest RMSE value of 3.16 which was obtained from a combination of C (0.05) and ε (1) that represented in TableI. In the experiment using GA, we got the lowest RMSE value of 3.09 which was obtained from a combination of optimum C (0.11544) and ε (0,518398) as shown as TableII.
The comparison graph of RMSE values in the SVM(dot) can be seen in the Fig2.
TABLE I. EXPERIMENT RESULT OF SVM(DOT)
C ε RMSE
0 0 3,307
0,1 0 3,149
0,01 0 3,12
0,001 0 3,263
0,01 0,1 3,121
-0,5 0,1 3,302
-0,5 0 3,307
-1 0 3,307
0,01 1 3,307
0,05 1 3,106
Mean 3,229
TABLE II. EXPERIMENT RESULT OF GA-SVM(DOT)
Range
Parameters
Optimum
Parameters RMSE
C ε C ε
-1 - 1 0 - 1 0,11601 0,07952 3,126
-1 - 1 0,1 - 1 0,11544 0,1459 3,118
-1 - 1 0,5 - 1 0,11544 0,518398 3,09
-1 - 1 0,8 - 1 0,11544 0,8 3,215
-0,5 - 1 0,5 -1 0,18386 0,72496 3,362
0,1 - 1 0,5 - 1 0,2495 0,9668 3,362
0-1 0,5-1 0,2495 0,9668 3,195
-1 - 1 0,3-1 0,11544 0,33215 3,113
-1 - 0,5 0,5 -1 -0,15224 0,53575 3,203
-1 - 0,5 1-10 -0,15224 1,62024 3,258
Mean 3,2042
Based on the experiment of SVM(dot), a statistical analysis of the different paired sample t-test was performed. The results obtained are listed in
TabelIII. Based on the different paired sample t-test, value of table t-stat is more than t critical two tail which means H1 is accepted, and the alpha value is less then 0,05 which means there is significant difference between SVM(dot) RMSE and GA-SVM(dot) RMSE.
TABLE III. STATISTIC DIFFERENT TEST RESULT
BETWEEN SVM(DOT) AND GA-SVM(DOT)
t-Test: Paired Two
Sample for Means
Variable 1
Variable
2
Mean 3,2289 3,2042
Variance 0,008433 0,009756
Observations 10 10
Pearson Correlation -0,07906 Hypothesized Mean
Difference 0
df 9
t Stat 0,55759
P(T<=t) one-tail 0,29536
t Critical one-tail 1,833113
P(T<=t) two-tail 0,590721
t Critical two-tail 2,262157
Fig2. RMSE Comparison of SVM(dot) and GA-SVM(dot)
In the polynomial kernel C and ε parameter are tried 10 times manually, and then we assign 10 minimum and maximum range value to the genetic algorithm to find the optimum parameters values to get the lowest RMSE. In the manual experiment, we got the lowest RMSE value of 3.085 which was obtained from a combination of C (0.1) ,ε (0) , and kernel degree (0,9) ,that represented in TableIV. In the experiment using GA, we got the lowest RMSE value of 3.06 which was obtained from a combination of optimum C (0.0659), ε (0,6828), and kernel degree (0,9345) as shown as TableV.
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10524
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative
Commons Attribution-NonCommercial 4.0 International License. 110
Based on the experiment of SVM(polynomial), a statistical analysis of the different paired sample t-test was performed. The results obtained are listed in TabelVI. Based on the different paired sample t-test, value of table t-stat is more than t critical two tail which means H1 is accepted, and the alpha value is less then 0,05 which means there is significant difference between SVM(polynomial) RMSE and GA-SVM(polynomial) RMSE.
TABLE IV. EXPERIMENT RESULT OF
SVM(POLYNOMIAL)
C ε Kernel
Degree
RMSE
0 0 2 3,437
0,05 0 2 3,496
0,01 0 2 3,453
0 0,01 2 3,453
0 0,1 2 3,453
0 0 1 3,095
0 0 0,5 3,095
0 0 0,1 3,095
0,1 0 0,1 3,088
0,1 0 0,9 3,085
Mean 3,275
TABLE V. EXPERIMENT RESULT OF GA-SVM(POLYNOMIAL)
Range Parameter Optimum Parameter
RMSE C ε Kernel
Degree
C ε Kernel
Degree
-1 -
1
0 -
1
0 – 1 -0,5493 0,51331 0,2378 3,043
-0,5 - 1
0,5 - 1
0 – 1 -0,4265 0,8567 0,9366 3,063
-1 - 1
0,5 - 1
0,5 - 1 -0,5493 0,5133 0,61754 3,043
-1 - 1
0 - 1
0 – 1 -0,88016
0,6883 0,9482 3,042
-1 - 1
0,01 - 1
0,01 - 1 -0,88016
0,688 0,94272 3,042
0 - 1 0,01 - 1
0 – 1 0,8633 0,6956 0,936637 3,049
0,01 - 1
0,01 - 1
0 – 1 0,0723 0,7136 0,9462 3,036
0,01 - 1
0,01 - 1
0 – 1 0,0874 0,6956 0,9366 3,060
0,01 - 1
0,01 - 1
0 – 1 0,0659 0,6828 0,9345 3,060
-0,1
- 1
0,01
- 1
0 – 1 0,7625 0,5633 0,63 3,068
Mean 3,0506
TABLE VI. STATISTIC DIFFERENT TEST RESULT
BETWEEN SVM(POLYNOMIAL) AND GA-SVM(POLYNOMIAL)
t-Test: Paired Two
Sample for Means
Variable
1
Variable
2
Mean 3,275 3,0506
Variance 0,037601 0,000124
Observations 10 10
Pearson Correlation -0,34764 Hypothesized Mean
Difference 0 df 9 t Stat 3,583039 P(T<=t) one-tail 0,002951 t Critical one-tail 1,833113 P(T<=t) two-tail 0,005903
t Critical two-tail 2,262157
The comparison graph of RMSE values in the
SVM(polynomial) can be seen in the Fig3.
Fig3. RMSE Comparison of SVM(Polynomial)
and GA-SVM(Polynomial)
In the Radial Basis Function kernel C and ε
parameter are tried 10 times manually, and then we assign 10 minimum and maximum range value to the
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10524
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative
Commons Attribution-NonCommercial 4.0 International License. 111
genetic algorithm to find the optimum parameters values to get the lowest RMSE. In the manual experiment, we got the lowest RMSE value of 2.957 which was obtained from a combination of C (0) ,ε (0.1) , and γ (0,1012) ,that represented in TableVII. In the experiment using GA, we got the lowest RMSE value of 2,92 which was obtained from a combination of optimum C (0.646), ε (0,8736), and γ(0,2634) as shown as TableVIII.
Experiment Result of SVM(RBF)
C ε γ RMSE
0 0 1 3,131
0,1 0 1 3,27
0 0,1 1 3,13
0 0,5 1 3,134
0 0,1 0,5 3,065
0 0,1 0,25 2,985
0 0,1 0,125 2,968
0 0,1 0,115 2,964
0 0,1 0,105 2,961
0 0,1 0,1012 2,957
Mean 3,057
Fig4. RMSE Comparison of SVM(RBF) and GA-SVM(RBF)
Based on the experiment of SVM(polynomial), a statistical analysis of the different paired sample t-test was performed. The results obtained are listed in TabelIX. Based on the different paired sample t-test, value of table t-stat is more than t critical two tail which means H1 is accepted, and the alpha value is less then 0,05 which means there is significant difference between SVM(RBF) RMSE and GA-SVM(RBF) RMSE. The comparison graph of RMSE values in the SVM(RBF) can be seen in the Fig4.
Based on the experiment of three kernel of SVM we can see that genetic algorithm absolutely can improve the RMSE performacne of SVM. The comparison between SVM and GA-SVM reprsented in Fig5.
TABLE VII. EXPERIMENT RESULT OF GA-SVM(RBF)
Range Parameter Optimum Parameter RMSE
C ε γ C ε γ
-1 -
1
0 -
1
0 - 1 -
0,5148
0,3984 0,193 2,957
-1 -
1
0,1-
1
0,1-
1
-
0,5224
0,4668 0,2766 2,972
-1 -
1
0 -
1
0,1-
1
-
0,5224
0,38014 0,2654 2,956
-1 -
1
0 -
1
0,01
- 1
-
0,5431
0,4164 0,1973 2,963
-1 -
1
0,01
- 1
0,1-
1
-
0,5501
0,3861 0,2654 2,965
0 -
1
0 -
1
0,1-
1
0,646 0,8736 0,2634 2,92
0 -
1
0,1-
1
0,1-
1
0,6034 0,893 0,28677 2,953
0 -
1
0 -
1
0,01
- 1
0,238 0,41146 0,18446 3,012
0 -
1
0 -
1
0 - 1 0,2654 0,9875 0,1885 2,97
0,1
- 1
0 -
1
0,1-
1
0,3095 0,4078 0,26197 2,999
Mean 2,9667
TABLE VIII. STATISTIC DIFFERENT TEST RESULT
BETWEEN SVM(RBF) AND GA-SVM(RBF)
t-Test: Paired Two Sample for
Means
Variable
1
Variable
2
Mean 3,0565 2,9667
Variance 0,011459 0,000636
Observations 10 10
Pearson Correlation -0,15826
Hypothesized Mean Difference 0
df 9
t Stat 2,495347
P(T<=t) one-tail 0,017061
t Critical one-tail 1,833113
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10524
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative
Commons Attribution-NonCommercial 4.0 International License. 112
P(T<=t) two-tail 0,034121
t Critical two-tail 2,262157
The last step of experiment is comparing GA-
SVM performance with another regression model, in
this case we use K-NN and Linear Regression. K-NN
and Linear Regression have been trained 10 times to
with each optimized parameter manually. K-NN
trained with differences of K value, hence Linear
Regression trained with differences of forward alpha
and backward alpha combination. The comparison of
the mean of RMSE value can be seen in Fig6.
Fig4. Meaan of RMSE Comparison of SVM) and GA-SVM
As we can see Fig4 shows that RBF kernel of SVM has lowest value event when SVM doesn’t improve by GA optimization.
Fig5. Meaan of RMSE Comparison of SVM) and GA-SVM
Based on Fig5. We can see that GA-SVM with RBF kernel has the lowest value of all regression models, the second rank is occupied by linear regression, and the third rank is GA-SVM with a polynomial kernel. All models show a slight difference in RMSE values. However, RMSE value
of the GA-SVM not at all close to zero, we suspect that the feature of this dataset must be selected first (Suryadi, 2019) like what another researcher do.
V. CONCLUSION AND SUGGESTION
Based on the entire experiment of the GA-SVM, we found that RBF kernel has a high performance of all regression model, however the RMSE generated in the GA-SVM experiment is still around number 2 and not at all close to zero which is a condition for a good RMSE value. In the case of the Liver Disorder dataset from BUPA, we suspect that the features of this dataset must be selected or weighed first, therefore for future work, we recommend selecting features or ranking features first in the BUPA dataset.
VI. ACKNOWLEDGMENT
A large debt of gratitude for “Way Back into Research” research group for the contribution of creative ideas around data mining research, and deep gratitude to RSW Institute research group which has been a source of inspiration from research that has been published so far.
VII. REFERENCES
Beeching, N., & Dassanayake, A. (2019). Tropical
liver disease. Medicine (United Kingdom),
47(11), 758–763.
https://doi.org/10.1016/j.mpmed.2019.08.001
Bhuvaneswari, P., & Therese, A. B. (2015). Detection
of Cancer in Lung With K-NN Classification
Using Genetic Algorithm. Procedia Materials
Science, 10(Cnt 2014), 433–440.
Chen, R., Liang, C.-Y., Hong, W.-C., & Gu, D.-X.
(2015). Forecasting holiday daily tourist flow
based on seasonal support vector regression
with adaptive genetic algorithm. Applied Soft
Computing, 26, 435–443.
https://doi.org/10.1016/j.asoc.2014.10.022
Goyal, R., Chandra, P., & Singh, Y. (2014).
Suitability of KNN Regression in the
Development of Interaction Based Software
Fault Prediction Models. IERI Procedia, 6, 15–
21. Retrieved from
http://dx.doi.org/10.1016/j.ieri.2014.03.004
Harafani, H., & Maulana, A. (2019). Penerapan
Algoritma Genetika pada Support Vector
Machine Sebagai Pengoptimasi Parameter
untuk Memprediksi Kesuburan. V(1), 51–59.
Harafani, H., & Wahono, R. S. (2015). Optimasi
Parameter pada Support Vector Machine
Berbasis Algoritma Genetika untuk Estimasi
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10524
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative
Commons Attribution-NonCommercial 4.0 International License. 113
Kebakaran Hutan. Journal of Intelligent
Systems, 1(2).
Kavousi-Fard, A., Samet, H., & Marzbani, F. (2014).
A new hybrid Modified Firefly Algorithm and
Support Vector Regression model for accurate
Short Term Load Forecasting. Expert Systems
with Applications, 41(13), 6047–6056.
https://doi.org/10.1016/j.eswa.2014.03.053
Kelly, D. A. (2019). Liver and biliary disease in
childhood. Medicine (United Kingdom),
47(12), 782–786.
https://doi.org/10.1016/j.mpmed.2019.09.011
Lantz, B. (2015). Machine Learning with R Learn.
Liang, H., Zou, J., Li, Z., Junaid, M., & Lu, Y. (2019).
Dynamic evaluation of drilling leakage risk
based on fuzzy theory and PSO-SVR
algorithm. Future Generation Computer
Systems, 95, 454–466.
https://doi.org/10.1016/j.future.2018.12.068
Lira, M. A. T., Da Silva, E. M., Alves, J. M. B., &
Veras, G. V. O. (2014). Estimation of wind
resources in the coast of Ceará, Brazil, using the
linear regression theory. Renewable and
Sustainable Energy Reviews, 39, 509–529.
Retrieved from
http://dx.doi.org/10.1016/j.rser.2014.07.097
Liu, H., Tian, H., Chen, C., & Li, Y. (2013). Electrical
Power and Energy Systems An experimental
investigation of two Wavelet-MLP hybrid
frameworks for wind speed prediction using
GA and PSO optimization. International
Journal of Electrical Power & Energy Systems,
52, 161–173.
https://doi.org/10.1016/j.ijepes.2013.03.034
Maimon, O., & Rokach, L. (2010). Data Mining and
Knowledge Discovery Handbook.
McDermott, J., & Forsyth, R. S. (2016). Diagnosing
a disorder in a classification benchmark.
Pattern Recognition Letters, 73, 41–43.
https://doi.org/10.1016/j.patrec.2016.01.004
Purwaningsih, E. (2019). Application of the Support
Vector Machine and Neural Network Model
Based on Particle Swarm Optimization for
Breast Cancer Prediction. SinkrOn, 4(1), 66.
https://doi.org/10.33395/sinkron.v4i1.10195
Raghavendra. N, S., & Deka, P. C. (2014). Support
vector machine applications in the field of
hydrology: A review. Applied Soft Computing,
19, 372–386. Retrieved from
http://linkinghub.elsevier.com/retrieve/pii/S15
68494614000611
Raikar, R. V., Wang, C. Y., Shih, H. P., & Hong, J.
H. (2016). Prediction of contraction scour using
ANN and GA. Flow Measurement and
Instrumentation, 50, 26–34.
https://doi.org/10.1016/j.flowmeasinst.2016.06
.006
Suryadi, A. (2019). Integration of Feature Selection
with Data Level Approach for Software Defect
Prediction. SinkrOn, 4(1), 51.
https://doi.org/10.33395/sinkron.v4i1.10137
Venkata Ramana, B., Babu, M. S. P., &
Venkateswarlu, N. . (2011). A Critical Study of
Selected Classification Algorithms for Liver
Disease Diagnosis. International Journal of
Database Management Systems, 3(2), 101–
114. https://doi.org/10.5121/ijdms.2011.3207
Wang, X., Wen, J., Zhang, Y., & Wang, Y. (2014).
Optik Real estate price forecasting based on
SVM optimized by PSO. Optik - International
Journal for Light and Electron Optics, 125(3),
1439–1443. Retrieved from
http://dx.doi.org/10.1016/j.ijleo.2013.09.017
Yao, Z., Li, J., Guan, Z., Ye, Y., & Chen, Y. (2020).
Liver disease screening based on densely
connected deep neural networks. Neural
Networks, 123, 299–304.
https://doi.org/10.1016/j.neunet.2019.11.005
Zaghloul, M. S., Hamza, R. A., Iorhemen, O. T., &
Tay, J. H. (2020). Comparison of adaptive
neuro-fuzzy inference systems (ANFIS) and
support vector regression (SVR) for data-
driven modelling of aerobic granular sludge
reactors. Biochemical Pharmacology,
(November 2019), 103742.
https://doi.org/10.1016/j.jece.2020.103742
Zhang, L., & Wang, N. (2018). Application of
coRNA-GA based RBF-NN to model proton
exchange membrane fuel cells. International
Journal of Hydrogen Energy, 43(1), 329–340.
https://doi.org/10.1016/j.ijhydene.2017.11.027
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10520
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative
Commons Attribution-NonCommercial 4.0 International License. 115
Simple Additive Weighting for Decision
Support Selection of Expedition Services Suhar Janti Mohammad Adriansyah Ghofar Taufik
Bina Sarana Informatika University STMIK Nusa Mandiri Bina Sarana Informatika University
Jakarta, Indonesia Jakarta, Indonesia Jakarta, Indonesia
[email protected] [email protected] [email protected]
Submitted: Mar 6, 2020
Accepted: Mar 29, 2020
Published: Apr 3, 2020
Abstract— Freight forwarding services are increasingly developing and each delivery
service provider competes to provide the best service, resulting in competition in terms
of price and delivery time, in order to attract the attention of users of shipping services.
The number of service providers with various types of packages offered by freight
forwarding services, making users difficult in determining the right service provider.
One way to overcome this problem is by the existence of a method that can provide
recommendations as consideration for making appropriate decisions. This study aims to
create a decision support system for the selection of goods delivery services by applying
the Simple Additive Weighting method that can solve problems by comparing between
shipping services. The results of this study are in the form of conclusion calculations
that can be taken into consideration for decision making in choosing the most widely
chosen freight forwarding services for students and getting the best results in decision
making. The results of calculations using the Simple Additive Weighting method, the
highest value based on time criteria is JNE YES with a value of 0.73, based on price
criteria is a vehicle with a value of 0.68, based on the weight criteria is JNE YES with a
value of 0.75, while based on the volume criteria the highest value is a vehicle with a
value of 0.70 .
Keywords—SAW; DSS; Expedition
I. INTRODUCTION
Freight forwarding services are increasingly
developing and each delivery service provider is
competing to provide the best service, resulting in
competition in terms of price and delivery time, in
order to attract the attention of users of shipping
services. The number of service providers with
various types of packages offered by freight
forwarding services, makes it difficult for users to
determine the right service provider with
considerations such as shipping prices, delivery time,
shipping weight and volume of goods sent.
To make the selection of shipping services in
shipping goods can use the Simple Additive
Weighting (SAW) method. Decision support systems
can be built as a tool for decision making in the
selection of expedition services. In this study the
criteria in selecting the expedition service are price,
time, weight and volume. Many previous studies that
use Simple Additive Weighting (SAW) which
discusses the decision support system in conducting
several alternatives with an assessment of existing
criteria, including:
In his research (Prihatin, Retnasari, & Fikri, 2019)
it can be concluded that PT. Buana Estate Bukit
Hambalang Villa Agrowisata in making decisions in
choosing its best employees using Simple Additive
Weighting (SAW). To facilitate decision making in
evaluating employees must determine priorities,
weights, or rankings based on the criteria given.
Specified criteria include attendance, performance,
discipline, attitude and neatness. This study got the
results of the number of employees as many as 25
employees who were used as research samples, the
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10520
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative
Commons Attribution-NonCommercial 4.0 International License. 116
employee named Pathurochman got the largest score
of 0.95 and got the best employee title at PT. Buana
Estate Bukit Hambalang Villa Agrotourism.
Other research (Ardhy & Efendi, 2019) is giving
rewards to employees using the Simple Additive
Weighting method by using several criteria consisting
of values of age, education, psychology, interviews,
work experience and health. Based on the weighting
calculation using the Simple Additive Weighting
(SAW) method, the employee with the highest value
is alternative A5 with a value of 85 and A6 with a
value of 72.5.
II. LITERATURE REVIEW
A. Decision Support System
Decision Support System is a computer-based
information system that approaches to produce
various alternative decisions to assist certain parties
in dealing with problems using data and models
(Nurjannah et al., 2015). Decision support systems
are designed to support all stages of decision making
ranging from identifying problems, selecting relevant
data, and determining the approach used in the
decision making process, to evaluating alternative
choices (Hidayat, Widiyanto, & Hasim, 2017).
Decision making is the result of a selection process of
various alternative actions that might be selected with
certain mechanisms, with the aim of producing the
best decision (Nurjannah et al., 2015)
Some characteristics contained in the decision
support system are (Badrul, Rusdiansyah, &
Budihartanti, 2019) :
a. Can support an organization or company in
making the decision process.
b. The existence of a human or interface (GUI) and
humans as users hold the control in carrying out a
process
c. Support decision making to discuss problems in a
structured manner and support interactions for
multiple decisions
d. Has a dialogue capacity to get information
according to needs
e. Having an integrated subsystem in such a way can
function as a system unity
f. Has two main components, namely data and mode
B. Multiple Attribute Decision Making
Multiple Attribute Decision Making (MADM) is
a method used to find optimal alternatives from a
number of alternatives with certain criteria. The
essence of FMADM is to determine the weight value
for each attribute, then proceed with a ranking
process that will select alternatives that have been
given (Daniati, 2015). Multiple Criteria Decision
Making (MCDM) is a decision making method to
determine the best alternative from a number of
alternatives based on certain criteria. Criteria are
usually in the form of measurements, rules or
standards used in decision making (Kusumadewi &
Purnomo, 2013). According to Zimmermann, that
based on its objectives, MCDM can be divided into
two models namely, Multi Attribute Decision Making
(MADM) and Multi Objective Decision Making
(MODM) (Kusumadewi & Purnomo, 2013).
III. PROPOSED METHOD
Simple Additive Weighting
One method of solving the MADM (Multiple
Attribute Decision Making) problem is to use the
Simple Additive Weighting method (Adianto, Arifin,
& Khairina, 2017). The basic concept of the SAW
method is to find a weighted sum of the performance
ratings for each alternative on all attributes (Manao et
al., 2017). The SAW method requires the process of
normalizing the decision matrix (x) to a scale that can
be compared with all available alternative ratings
(Ikhmah & Widawati, 2018).
In the Simple Additive Weighting (SAW)
method, it is grouped into various criteria, then
verification of fuzzy numbers in the form of crisp
numbers so that the values will be carried out in the
calculation process to find the best alternative (Gani,
Kridalaksana, & Arifin, 2019). In the Simple
Additive Weighting (SAW) method of calculating
alternative total scores, the assessment of each
attribute must pass through normalization first. The
process of normalizing the decision matrix (x) to
which scale can be compared with all alternative
assessments carried out by the following formula
(Ketaren, 2016) :
𝑟𝑖𝑗 = {
𝑥𝑖𝑗𝑚𝑎𝑥 𝑥𝑖𝑗
,
𝑚𝑖𝑛𝑥𝑖𝑗𝑥𝑖𝑗
,
Information :
1. The symbol rij is the normalized performance
rating of alternative Ai on the attributes Cj, i = 1,2,
..., m and j = 1,2, ..., n. (m and n are the many
alternatives and criteria).
2. The xij symbol is the matching rating value on Ai
and Cj.
3. The max xij symbol is the largest value of all match
rating values for each criterion.
4. Min xij symbol is the smallest value of all match
ratings for each criterion.
if j is the profit attribute (benefit)
if j is the cost attribute (cost)
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10520
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative
Commons Attribution-NonCommercial 4.0 International License. 117
5. The profit attribute is if the greatest value in the
attribute is the best value.
6. The cost attribute is if the smallest value in the
attribute is the best value.
The preference value for each alternative (Vi) is given
as follows:
𝑉𝑖 =∑𝑤𝑗𝑟𝑖𝑗
𝑛
𝑗=1
Information :
1. The symbol Vi is the ranking for each alternative
2. The symbol n is the number (number) of
alternatives
3. The wj symbol is the weight value of each criterion
4. The rij symbol is a normalized performance rating
value
A greater value of Vi indicates that the alternative Ai
is preferred.
The alternative management steps that are used (in
this case the decision to select shipping services),
include (Surya & Wahyu, 2020) :
1. Determine the alternative, namely Ai,
2. Determine the criteria that will be used as a
reference in making decisions, namely Cj,
3. Determine the weight of preference or level of
importance (W) of each criterion
W=[W1, W2, W3, W4, …, Wj]
4. Give a rating of the suitability of each alternative
on each criterion,
5. Make a decision matrix (x) formed from the
match rating table of each alternative to each
criterion, the value of x every alternative (Ai) to
each predetermined criterion (Cj), where i =
1,2, ..., m and j = 1,2, ..., n
= [
𝑥11 𝑥12 𝑥21 𝑥22
⋯ 𝑥1𝑛 … 𝑥2𝑛
⋮ ⋮ ⋮ 𝑥𝑚1 𝑥𝑚2 ⋯ 𝑥𝑚𝑛
]
6. Normalizing the decision matrix (x) to a scale that
can be compared with all available alternative
ratings.
𝑅𝑖𝑗 =
𝑋𝑖𝑗
𝑀𝑎𝑥 𝑋𝑖𝑗
7. The results of matrix normalization (Rij) form a
normalized matrix (R).
𝑅 = [
𝑟11 𝑟12 𝑟21 𝑟22
⋯ 𝑟1𝑛 … 𝑟2𝑛
⋮ ⋮ ⋮ 𝑟𝑚1 𝑟𝑚2 ⋯ 𝑟𝑚𝑛
]
8. The end result of the preference value (Vi) is
obtained from the sum of the multiplications of
normalized matrix row elements (R) with
preference weights (W) corresponding to the
matrix column elements (R).
9. The ranking process is obtained based on the
alternative which has the largest to the lowest total
value as the best alternative.
IV. RESULT AND DISCUSSION
A. Requirement Analysis
The needs analysis in this case is divided into two
parts, namely the input requirements analysis and the
output needs analysis. The input needs analysis used
is the following variables (criteria):
1. Price
2. Time
3. Weight
4. Volume
From these variables (criteria), the level of
importance is determined based on the weighted
value to the fuzzy numbers. Alternative match ratings
for the following criteria:
TABLE I. FUZZY NUMBERS
Fuzzy Number Score
Very Low 1
Low 2
Satis 3
Hight 4
Very High 5
Based on the criteria and rating match, alternative
(Ai) to the criteria (Cj), then the translation of the
criteria weights (Cj) are converted to Fuzzy numbers.
As for the analysis of output needs in this study is an
alternative that has the highest value compared to
other value alternatives. What is meant by
alternatives are the types of freight forwarding
services mentioned earlier.
B. Data Processing and Calculation With SAW
At this data processing stage, calculation or
testing of the data that has been presented will be
carried out using the Simple Additive Weighting
(SAW) method which is carried out for data
processing.
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10520
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative
Commons Attribution-NonCommercial 4.0 International License. 118
There are several steps to calculate the selection
of goods shipping services using the simple additive
weighting method, including:
1. Alternative Data Delivery Services
The first step is to determine the alternative Ai, here
are the alternative data that will be used in the
calculation.
TABLE II. ALTERNATIVE DATA
No Name Price
(IDR) Time
Weight
(Max) Volume
1 JNE YES 18,000 one
day 150 Kg 6,000
2 JNE
Reguler 9,000
1-2
days 150 Kg 6,000
3 J&T 10,000 1-3
days 100 Kg 7,600
4 Wahana 5,000 1-4
days 50 Kg 6,000
5 Tiki Reg 9,000 1-7
days 75 Kg 6,000
6 Tiki Ons 17,000 1-1
days 75 Kg 6,000
2. Criteria and Weight
The second step determines the criteria that will be
used as a reference in decision making (Cj),
including:
a. Price
Values and Weights for prices are shown below:
TABLE III. PRICE
Price Range
(IDR) Fuzzy Number Score
1,000 - 5,000 Very Low 1
6,000 - 10,000 Low 2
11,000 - 15,000 Satis 3
16,000 - 20,000 Hight 4
> 20,000 Very High 5
b. Time
Value and Weight for delivery time are shown:
TABLE IV. TIME
Time Fuzzy Number Score
1 day (max) Very Low 1
2 days (max) Low 2
3 days (max) Satis 3
4 days (max) Hight 4
>5 days Very High 5
c. Weight
Value and Weight for weight are:
TABLE V. WEIGHT (MAX)
Weight (Max) Fuzzy Number Score
1 - 25 kg Very Low 1
26 - 50 kg Low 2
51 - 100 kg Satis 3
101 - 150 kg Hight 4
> 150 kg Very High 5
d. Volume
Values and Weights for volumes are shown below:
TABLE VI. VOLUME
Volume (cm) Fuzzy Number Score
1 - 2,500 Very Low 1
2,500 - 5,000 Low 2
5,000 - 7,500 Satis 3
7,500 - 1,000 Hight 4
>10,000 Very High 5
3. Weight Preferences (W)
The third step is to determine the preference
weights or the importance level (W) of each criterion.
The weight of this criterion that is used in selecting
goods shipping services is as follows:
TABLE VII. IMPORTANCE (W)
Criteria (Cj) Information Weight (W)
(C1) Price Satis 30%
(C2) Time Hight 40%
(C3) Weight Very Low 10%
(C4) Volume Low 20%
4. Match rating value of each alternative on each
criterion
The fourth step is to determine the suitability
rating of each alternative on each of the criteria
specified above indicated:
TABLE VIII. ALTERNATIVE MATCH RATINGS
Alternative
(Delivery
service)
Criterias
Price
(C1)
Time
(C2)
Weight
(C3)
Volume
(C4)
A1 4 1 4 3
A2 2 2 4 3
A3 2 3 3 4
A4 1 4 2 3
A5 2 5 3 3
A6 4 1 3 3
5. Matrix of Decisions
After alternative rating values for each criterion are
determined, the fifth step is to make a decision matrix
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10520
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative
Commons Attribution-NonCommercial 4.0 International License. 119
(x) formed from the match rating table of each
alternative for each criterion, the value of x every
alternative (Ai) for each predetermined criterion (Cj).
(
4 1 4 32 2 4 32 3 3 41 4 2 32 5 3 34 1 3 3)
6. Normalization of Decision Matrix (x)
The sixth step is to normalize the decision matrix
(x) to a scale that can be compared with all existing
alternative ratings. The value provided is a match
value. For matrices R11 through R62 calculated using
the formula:
𝑅𝑖𝑗 = 𝑀𝑖𝑛 𝑥𝑖𝑗
𝑥𝑖𝑗
With the example calculation as follows:
𝑅11 = 𝑀𝑖𝑛 (4; 2; 2; 1; 2; 4)
4= 1
4= 0.25
𝑅21 = 𝑀𝑖𝑛 (4; 2; 2; 1; 2; 4)
2= 1
2= 0.50
For matrices R13 through R64 calculated by the
formula
𝑅𝑖𝑗 = 𝑥𝑖𝑗
𝑀𝑎𝑥 𝑥𝑖𝑗
With the example calculation as follows:
𝑅13 = 4
𝑀𝑎𝑥 (4;4; 3; 2; 3; 3)= 4
4= 1
𝑅33 = 3
𝑀𝑎𝑥 (4;4; 3; 2; 3; 3)= 3
4= 0.75
7. Normalized Matrix (R)
The results of matrix normalization (Rij) form a
normalized matrix (R). And here is the normalized
(R) matrix data.
(
0.25 1.00 1.00 0.750.50 0.50 1.00 0.750.50 0.33 0.75 1.001.00 0.25 0.50 0.750.50 0.20 0.75 0.750.25 1.00 0.75 0.75)
8. Value Preference (Vi)
The next step is to calculate the final preference
value (Vi) obtained from the calculation of the
normalized matrix row element (R) with the
preference weight (W) corresponding to the matrix
column element (R):
a. Time
Weight for time criteria is
TABLE IX. WEIGHTS(W) OF TIME
Weights(W) of Time
Price
(C1)
Time
(C2)
Weight
(C3)
Volume
(C4) Total
30% 40% 10% 20% 100
The calculation is as follows:
V1 = [(0,3 x 0,25) + (0,4 x 1,00) + (0,1 x1,00) + (0,2
x 0,75) ] = 0,73
V2 = [(0,3 x 0,50) + (0,4 x 0,50) + (0,1 x1,00) + (0,2
x 0,75) ] = 0,60
V3 = [(0,3 x 0,50) + (0,4 x 0,33) + (0,1 x0,75) + (0,2
x 1,00) ] = 0,56
V4 = [(0,3 x 1,00) + (0,4 x 0,25) + (0,1 x0,50) + (0,2
x 0,75) ] = 0,60
V5 = [(0,3 x 0,50) + (0,4 x 0,20) + (0,1 x0,75) + (0,2
x 0,75) ] = 0,46
V6 = [(0,3 x 0,25) + (0,4 x 1,00) + (0,1 x0,75) + (0,2
x 0,75) ] = 0,70
The following data are the results of the calculation
of preference values presented in
TABLE X. PREFERENCE VALUE
CALCULATION RESULTS
Alternative
(Delivery
service)
Criterias
Outcome Price
(C1)
Time
(C2)
Weight
(C3)
Volume
(C4)
A1 4 1 4 3 0,73
A2 2 2 4 3 0,60
A3 2 3 3 4 0,56
A4 1 4 2 3 0,60
A5 2 5 3 3 0,46
A6 4 1 3 3 0,70
The biggest value is in V1, so alternative A1 (JNE
YES) is the alternative chosen as the best alternative
with the final result = 0.73.
b. Price
Weights for prices can be seen as follows
TABLE IX. WEIGHTS(W) OF PRICE
Weights(W) of Price
Price
(C1)
Time
(C2)
Weight
(C3)
Volume
(C4) Total
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10520
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative
Commons Attribution-NonCommercial 4.0 International License. 120
40% 30% 10% 20% 100
The calculation is as follows:
V1 = [(0,4 x 0,25) + (0,3 x 1,00) + (0,1 x1,00) + (0,2
x 0,75) ] = 0,65
V2 = [(0,4 x 0,50) + (0,3 x 0,50) + (0,1 x1,00) + (0,2
x 0,75) ] = 0,60
V3 = [(0,4 x 0,50) + (0,3 x 0,33) + (0,1 x0,75) + (0,2
x 1,00) ] = 0,58
V4 = [(0,4 x 1,00) + (0,3 x 0,25) + (0,1 x0,50) + (0,2
x 0,75) ] = 0,68
V5 = [(0,3 x 0,50) + (0,4 x 0,20) + (0,1 x0,75) + (0,2
x 0,75) ] = 0,49
V6 = [(0,3 x 0,25) + (0,4 x 1,00) + (0,1 x0,75) + (0,2
x 0,75) ] = 0,63
The following data are the results of the calculation
of preference values presented in
TABLE XI. PREFERENCE VALUE
CALCULATION RESULTS
Alternative
(Delivery
service)
Criterias Out
come Price
(C1)
Time
(C2)
Weight
(C3)
Volume
(C4)
A1 4 1 4 3 0,65
A2 2 2 4 3 0,60
A3 2 3 3 4 0,55
A4 1 4 2 3 0,68
A5 2 5 3 3 0,49
A6 4 1 3 3 0,63
The greatest value is in V4, so that alternative A4
(Wahana) is the alternative chosen as the best
alternative with the final result = 0.68.
c. Weight
Weight for weight criteria is
TABLE XII. WEIGHTS(W) OF WEIGHT
Weights(W) of Weight
Price
(C1)
Time
(C2)
Weight
(C3)
Volume
(C4) Total
30% 20% 40% 10% 100
The calculation is
V1 = [(0,3 x 0,25) + (0,2 x 1,00) + (0,4 x1,00) + (0,1
x 0,75) ] = 0,75
V2 = [(0,3 x 0,50) + (0,2 x 0,50) + (0,4 x1,00) + (0,1
x 0,75) ] = 0,73
V3 = [(0,3 x 0,50) + (0,2 x 0,33) + (0,4 x0,75) + (0,1
x 1,00) ] = 0,62
V4 = [(0,3 x 1,00) + (0,2 x 0,25) + (0,4 x0,50) + (0,1
x 0,75) ] = 0,63
V5 = [(0,3 x 0,50) + (0,2 x 0,20) + (0,4 x0,75) + (0,1
x 0,75) ] = 0,57
V6 = [(0,3 x 0,25) + (0,2 x 1,00) + (0,4 x0,75) + (0,1
x 0,75) ] = 0,65
The following data are the results of the calculation
of preference values presented in
TABLE XIII. PREFERENCE VALUE
CALCULATION RESULTS
Alternative
(Delivery
service)
Criterias Out
come Price
(C1)
Time
(C2)
Weight
(C3)
Volume
(C4)
A1 4 1 4 3 0,75
A2 2 2 4 3 0,73
A3 2 3 3 4 0,62
A4 1 4 2 3 0,63
A5 2 5 3 3 0,57
A6 4 1 3 3 0,65
The biggest value is in V1, so alternative A1 (JNE
YES) is the alternative chosen as the best alternative
with the final result = 0.75.
d. Volume
Weight for volume criteria is
TABLE XIV. WEIGHTS(W) OF VOLUME
Weights(W) of Weight
Price
(C1)
Time
(C2)
Weight
(C3)
Volume
(C4) Total
30% 20% 10% 40% 100
The calculation is
V1 = [(0,3 x 0,25) + (0,2 x 1,00) + (0,1 x1,00) + (0,4
x 0,75) ] = 0,68
V2 = [(0,3 x 0,50) + (0,2 x 0,50) + (0,1 x1,00) + (0,4
x 0,75) ] = 0,65
V3 = [(0,3 x 0,50) + (0,2 x 0,33) + (0,1 x0,75) + (0,4
x 1,00) ] = 0,69
V4 = [(0,3 x 1,00) + (0,2 x 0,25) + (0,1 x0,50) + (0,4
x 0,75) ] = 0,70
V5 = [(0,3 x 0,50) + (0,2 x 0,20) + (0,1 x0,75) + (0,4
x 0,75) ] = 0,57
V6 = [(0,3 x 0,25) + (0,2 x 1,00) + (0,1 x0,75) + (0,4
x 0,75) ] = 0,65
The following data are the results of the calculation
of preference values presented in
TABLE XV. PREFERENCE VALUE
CALCULATION RESULTS
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10520
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative
Commons Attribution-NonCommercial 4.0 International License. 121
Alternative
(Delivery
service)
Criterias Out
come Price
(C1)
Time
(C2)
Weight
(C3)
Volume
(C4)
A1 4 1 4 3 0,68
A2 2 2 4 3 0,65
A3 2 3 3 4 0,69
A4 1 4 2 3 0,70
A5 2 5 3 3 0,57
A6 4 1 3 3 0,65
The biggest value is in V4, so the A4 alternative
(Wahana) is the alternative chosen as the best
alternative with the final result = 0.70.
V. CONCLUSION AND SUGGESTION
Simple Additive Weighting method can help
produce the best alternative decisions in the decision
of the choice of goods shipping services. And these
alternatives are in accordance with the criteria that
influence the selection of freight forwarding services.
From the process of calculating and ranking the
alternatives obtained the highest value results which
are the results required for consideration and
recommendations for the user or in this case the users
of goods delivery services. And the best alternative
based on time is JNE YES, based on price is vehicle,
based on weight is JNE YES and based on volume is
vehicle.
The results of calculations using the Simple
Additive Weighting method, the highest value based
on time criteria is JNE YES with a value of 0.73,
based on price criteria is a vehicle with a value of
0.68, based on the weight criteria is JNE YES with a
value of 0.75, while based on the volume criteria the
highest value is a vehicle with a value of 0.70.
VI. REFERENCES
Adianto, T. R., Arifin, Z., & Khairina, D. M. (2017).
Sistem Pendukung Keputusan Pemilihan
Rumah Tinggal Di Perumahan Menggunakan
Metode Simple Additive Weighting (Saw)
(Studi Kasus : Kota Samarinda). In Prosiding
Seminar Ilmu Komputer dan Teknologi
Informasi (Vol. 2, pp. 197–201).
Ardhy, F., & Efendi, D. M. (2019). Pemberian
Reward Terhadap Karyawan Terbaik Dengan
Menggunakan Metode Simple Additive
Weighting ( SAW ). Jurnal Sistem Informasi &
Manajemen Basis Data (SIMADA), 2(2), 176–
181.
Badrul, M., Rusdiansyah, R., & Budihartanti, C.
(2019). Application of Simple Additive
Weighting Method for Determination of
Toddler Nutrition Status. SinkrOn, 4(1), 19.
https://doi.org/10.33395/sinkron.v4i1.10145
Daniati, E. (2015). Sistem Pendukung Keputusan
Pemilihan Kost Di Sekitar Kampus UNP Kediri
Menggunak Metode Simple Additive
Weighting (SAW). Seminar Nasional
Teknologi Informasi Dan Multimedia 2015,
(Pemilihan Kost), 2.2-145-2.2-150.
Gani, A., Kridalaksana, A. H., & Arifin, Z. (2019).
Analisa Perbandingan Metode Simple Additive
Weighting (SAW) Dan Weight Product (WP)
Dalam Pemilihan Kamera Mirrorless. Jurnal
Ilmiah Ilmu Komputer, 14(2), 76–81.
Hidayat, T., Widiyanto, F., & Hasim, Y. K. (2017).
Rancang Bangun Decision Support System
Pemilihan Guru Terbaik Menggunakan Metode
Simple Additive Weighting (Saw)(Studi ….
JUTIS Journal of Informatics Engineering,
5(1), 52–56. Retrieved from
http://ejournal.unis.ac.id/index.php/jutis/article
/view/5
Ikhmah, & Widawati, A. S. (2018). Sistem
Pendukung Keputusan Pemilihan Tempat
Wisata Purworejomenggunakan Metode Saw.
Seminar Nasional Teknologi Informasi Dan
Multimedia 2018, 91–96.
Ketaren, E. (2016). Utility Vectors To Fuzzy
Preference Relation Dengan Metode Simple
Additive Weighting ( SAW ) Dalam Penentuan
Posisi Kerja Karyawan. SinkrOn, 1(1), 6–9.
Retrieved from
https://jurnal.polgan.ac.id/index.php/sinkron/ar
ticle/view/2
Kusumadewi, S., & Purnomo, H. (2013). Aplikasi
Logika Fuzzy Untuk Pendukung Keputusan
(2nd ed.). Yogyakarta: Graha Ilmu.
Manao, H., Nadeak, B., & Zebua, T. (2017). Sistem
Pendukung Keputusan Pemilihan Perumahan
Dengan Metode Simple Additive Weighting
(Saw). Media Informatika Budi Darma, 1(2),
49–53.
Nurjannah, N., Arifin, Z., & Khairina, D. M. (2015).
Sistem Pendukung Keputusan Pembelian
Sepeda Motor Dengan Metode Weighted
Product. Informatika Mulawarman : Jurnal
Ilmiah Ilmu Komputer, 10(2), 20.
https://doi.org/10.30872/jim.v10i2.186
Prihatin, T., Retnasari, T., & Fikri, M. (2019). A
Determination of The Best Employees using
Simple Additive Weighting (SAW) Method.
SinkrOn, 4(1), 106.
https://doi.org/10.33395/sinkron.v4i1.10169
Surya, C., & Wahyu, A. (2020). Sistem Informasi
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10520
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative
Commons Attribution-NonCommercial 4.0 International License. 122
Perhitungan Poin Pelanggaran Siswa
Menggunakan Metode Simple Additive
Weighting (SAW) (Studi Kasus Di SMK As-
Shofa Kabupaten Tasikmalaya). Jurnal
TEKNOINFO, 14(1), 59.
https://doi.org/10.33365/jti.v14i1.477
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10528
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative
Commons Attribution-NonCommercial 4.0 International License. 123
Analysis of Taklinear Performance and Integer
Linear Programming Models in Nurses
Scheduling Problems
Junerdi Nababan
University Sumatera Utara
Medan, Indonesia
Tulus
University Sumatera Utara
Medan, Indonesia
Zakarias Situmorang
University Sumatera Utara
Medan, Indonesia
Submitted: Mar 13, 2020
Accepted: Mar 26, 2020
Published: Apr 6, 2020
Abstract— Almost all hospitals schedule nurses' work shifts manually, which is not
effective, mistakes in scheduling nurses in hospitals can make them not work optimally
so that they are prone to making mistakes and will endanger every patient in the hospital.
In the research will be proposed integer linear programming model and branch and
bound method, the purpose of this study is to develop an optimization model of nurse
scheduling problems at the hospital. The optimization model will minimize the total
deviation of nurses' working days from the standard workdays, the optimization model
implemented in hospitals is processed using LINGO software.
Keywords— Nursing Scheduling Issues, Integer Linear Programming, LINGO.
I. INTRODUCTION
The problem of nurse scheduling is an important issue that must be considered in every hospital, good nurse scheduling must be applied in every hospital so that ongoing activities are not interrupted and more optimal. Errors in arranging the scheduling of nurses at the hospital can make them work less optimally so they are prone to making mistakes and will endanger every patient in the hospital. Nurse scheduling problem or known as Nurse Scheduling Problem (NSP) (Hakim, Bakhtiar, & Jaharuddin, 2017) (Jafari & Salmasi, 2015), NSP is a nurse scheduling problem at the hospital in arranging the shift scheduling for each nurse. Nursing scheduling problems at hospitals often occur because many hospitals do manual scheduling. Scheduling nurses manually is ineffective and inefficient wherein arranging the scheduling must pay attention to the limited number of nurses that exist and must pay attention to the uniformity of the schedule given to each nurse.
In this study, the authors tried to solve the problem of scheduling nurses applying the Integer Linear Programming model and the branch and bound method, then processed using the LINGO application. Branch and bound algorithm is one method that can be used to solve Integer Programming cases. This method divides the problem into sub-problems that lead to solutions by forming a search tree structure and limiting it to achieve an optimal solution (Sumathi, 2016) (Bakhtiar & Jaharuddin, 2017). Branch and Bound algorithm procedures are carried out repeatedly to form a search tree and bounding process by determining the upper bound and lower bound in finding the optimal solution (Suryawan, Tastrawati, & Sari, 2016)
Literature Review
(Angeline, Iryanto, & Tarigan, 2014) apply the branch and bound method in determining the optimum amount of production at Cv. Xyz, in his research, reviewed based on the amount of raw material inventory, market demand, profits, and time
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10528
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative
Commons Attribution-NonCommercial 4.0 International License. 124
of manufacture of each pant. The results showed that the optimal amount of production of each type of pants (men's trousers, women's trousers, men's shorts, women's shorts). (Nur & Abdal, 2016) Using the branch and bound and memory cut methods in determining integer linear programming solutions, the results of the Branch and Bound and Gomory Cut methods can be used to solve linear integer programming problems. Based on the case examples given, the same results were obtained between the Branch and Bound and Gomory Cut methods, where the company had to produce 47 A flashlights, 62 B flashlights, and 172 C flashlights in order to produce an optimum profit of 2,233,000. In the process of completion, the branch and bound method require a large number of simplex iterations and a log time but more convergence is guaranteed. (Suryawan, Tastrawati, & Sari, 2016) Applying the branch and bound algorithm method in optimizing bread production results in a 25.2% increase in profits through calculations by applying the branch and bound algorithm. (Rafeek & Siswanto, 2015) Make an ITS course schedule and then develop an integer programming (ILP) method to solve NP-Complete problems using LINGO software. (Hasan & Arefin, 2017)
II. PROPOSED METHOD
1. Datasheet
The data used in this study were taken from Cut Nyak Dien Tapak Tuan Aceh Selatan Hospital, taking into account the rules set at the hospital, each nurse would be scheduled with a fair amount of work and shift where all nurses who worked no more than 22 days work, nurses work no more than one shift a day, each nurse works no more than 6 days a week, each nurse works no more than 48 hours a week, if every nurse on duty at the night shift should not be followed by a morning shift the next day, the nurse's schedule must meet the minimum number of nurses needed in each shift every day, nurses who work on the night shift for 2 consecutive days in one week then the next day is given a day off, and each nurse works no more than 2-night shift in one week.
2. Modeling
After the stage of formulating the problem, the next step is to represent the problem in the mathematical model. Through this model, the problem is described as a system of equations or other inequalities and mathematical expressions. Nurse work scheduling problems can be modeled as Integer Linear Programming (ILP) (KOÇ & AKTAN, 2019).
3. Model solution
A mathematical model for solving real problems requires computer performance. To get the scheduling model solution, LINGO software is used with the branch and bound method. Then the solution obtained is the solution that most accommodates all restrictions and minimizes the objective function.
The definition of notation that will be used in this model is
Index:
i = Nurse (i = 1, 2, …, I).
j = Shift (j = 1, …, J).
1 = Morning Shift (pukul 07.00-14.00).
2 = Afternoon Shift (pukul 14.00-21.00).
3 = Night Shift (pukul 21.00-07.00).
k = Day (k = 1, 2, …, K).
T = Total working days in the one month assignment period.
a = The time constant in the day for one-week assignment.
b = The time constant in hours for one week of assignment.
c = Constant length of work time for each shift.
djk = The number of nurses working for each shift j on day k.
Variabel keputusan:
jika perawat i bekerja di shift j pada
hari k. selainnya
𝑦1𝑖 = Deviation of the shortage of workdays for each nurse i (slack variable)
𝑦2𝑖 = Deviation of excess workdays for each nurse i (variable surplus).
Fungsi tujuan adalah meminimumkan total deviasi dari kendala penyimpangan hari kerja perawat.
Min 𝑧 = ∑ 𝑦1𝑖 + 𝑦2𝑖
𝑙
𝑖=1
Constraints:: 1. Total of the number of workdays per nurse
and deviation of workdays in one month T days,
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10528
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative
Commons Attribution-NonCommercial 4.0 International License. 125
(∑ ∑ 𝑥𝑖𝑗𝑘
𝐾
𝑘=1
𝐽
𝑗=1
) + 𝑦1𝑖 − 𝑦2𝑖 = 𝑇 ∀𝑖 = 1, … , 𝐼
2. Each nurse works no more than one shift every day,
∑ 𝑥𝑖𝑗𝑘 ≤
𝐽
𝑗=1
1 ∀𝑖 = 1, … , 𝐼, ∀𝑘 = 1, … , 24
3. Each nurse works no more than a day a week,
∑ ∑ 𝑥𝑖𝑗(𝑘−𝑡+7) ≤ 𝑎 ∀𝑖 = 1, … , 𝐼, ∀𝑘
7
𝑡=1
𝐽
𝑗=1
= 1, … . , 24
4. Each nurse works no more than b hours a week with long hours of work c hours per shift t,
∑ ∑ 𝑐. 𝑥𝑖𝑗(𝑘−𝑡+7) ≤ 𝑏 ∀𝑖 = 1, … , 𝐼, ∀𝑘
7
𝑡=1
𝐽
𝑗=1
= 1, … . , 24
5. Every nurse on duty at the night shift may not be followed by the morning shift the next day.
𝑥𝑖3𝑘 + 𝑥𝑙1(𝑘+1) ≤ 1 ∀𝑖 = 1, … , 𝐼, ∀𝑘
= 1, … . , 𝐾
6. The nurse schedules must meet the need for a minimum number of nurses in each shift each day,
∑ 𝑥𝑖𝑗𝑘 ≥ 𝑑𝑗𝑘 ∀𝑗 = 1, … , 𝐽, ∀𝑘
𝐼
𝑖=1
= 1, … . , 𝐾
7. If each nurse is on night shifts for two consecutive days, then the next day off,
𝑥𝑖3𝑘 + 𝑥𝑖3(𝑘+𝑙) + ∑ 𝑥𝑖𝑗(𝑘+2) ≤ 2 ∀𝑖
3
𝑗=1
= 1, … , 𝐼, ∀𝑘 = 1, … . , 𝐾
8. Each nurse works no more than 2-night shifts in one week,
∑ 𝑥𝑖3(𝑘−𝑙+7) ≤ 2 ∀𝑖 = 1, … , 𝐼, ∀𝑘
7
𝑡=1
= 1, … . , 24
9. The head nurse gets a vacation every Sunday,
𝑥1𝑗(7𝑘−𝑑+1) = 0 ∀𝑗 = 1, … , 𝐽, ∀𝑘
= 1, … . ,4, 𝑑 = 1, … , 7
10. Limitation of negative and integer.
𝑥𝑖𝑗𝑘 ∈ {0,1} ∀𝑖 = 1, … , 𝐼, ∀𝑘
= 1, … . , 𝐽, ∀𝑘 = 1, … , 𝐾
III. RESULT AND DISCUSSION
1. Nurse Scheduling Results in several Rooms.
After the mathematical model is formulated with the form of Integer Linear Programming, then processed using LINGO with the branch and bound method the nurses work schedule is generated for the emergency room nursing room, class, and ward at the General Hospital in one month with a minimum deviation of working days. Nurses' work schedules in each part of the room from the modeling results provide information on comparing working day deviations between manual scheduling and the new scheduling in September, as well as the total workdays of each nurse on morning, evening and night shifts. The need for a minimum number of nurses in each part of the room in the General Hospital from the modeling results is in accordance with the number of nurses needed. Working days and holidays for each nurse has fulfilled the hospital management's requirements.
Implementation of the model that has been obtained is done by means of model simulation. The simulation uses data on the number of nurses and the need for the number of nurses every day in each section of the room at the General Hospital.
The schedule arranged meets the rules set, namely:
All nurses who work do not deviate from 22 working days.
Nurses work no more than one shift in one day.
Each nurse works no more than 6 days a week.
Each nurse works no more than 48 hours a week.
If every nurse on duty at night shift then it should not be followed by the morning shift the next day.
The nurse schedules must meet the need for a minimum number of nurses in each shift each day
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10528
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative
Commons Attribution-NonCommercial 4.0 International License. 126
Nurses who work on the night shift for 2 consecutive days in one week then the next day is given a day off.
Each nurse works no more than 2-night shifts in one week.
There are nurses who cannot be scheduled on certain days. The head nurse is given a day off on Sundays.
The results of mathematical modeling formulated in the form of Integer Linear Programming and processed using LINGO software using the branch and bound method can be seen in the table below:
Table 1. Comparison of Deviations from Emergency Room Nurse Workdays
In table 1. it can be seen that a comparison of the scheduling model between the manual model and the new scheduling model in September shows that the new scheduling model gives all nurses a more equitable number of working days. In the new scheduling model, there are no deviations from workdays compared to manual scheduling which there are deviations of 6 workdays with excess 5 days and a deficiency of 1 workday can be seen in Table 1, The new schedule considers holidays, night shifts and workday needs of each nurse, so that the new schedule has paid attention to the nurses fatigue factor.
Table 2. Comparison of Classroom Nurse Workday Deviations.
Manual Schedule Emergency Room
Nurse Mor
ning
After
noon
Nig
ht
Total KB KK
1 11 11 1 22 1 0
2 8 8 7 23 1 0
3 7 8 7 22 0 0
4 6 9 7 22 0 0
5 6 9 6 21 0 1
6 7 7 9 23 1 0
7 6 8 8 22 0 0
8 10 6 7 23 1 0
9 10 4 8 22 0 0
10 6 10 7 23 1 0
11 5 10 7 22 0 0
12 8 7 7 22 0 0
13 7 6 9 22 0 0
14 9 7 6 22 0 0
Amount 5 1
Total Deviations 6
New Model Emergency Room
Nurse Mor
ning
Aftern
oon
Nig
ht
Total KB KK
1 8 9 5 22 0 0
2 11 4 7 22 0 0
3 7 8 7 22 0 0
4 9 6 7 22 0 0
5 10 7 5 22 0 0
6 8 7 7 22 0 0
7 10 7 5 22 0 0
8 14 2 6 22 0 0
9 10 6 6 22 0 0
10 8 7 7 22 0 0
11 9 5 8 22 0 0
12 7 8 7 22 0 0
13 8 7 7 22 0 0
14 4 12 7 22 0 0
Amount 0 0
Total Deviations 0
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10528
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative
Commons Attribution-NonCommercial 4.0 International License. 127
In table 2. it can be seen that a comparison of scheduling models between manual models and the new scheduling model in September shows that the new scheduling model provides a more equal number of working days to all nurses. In the new scheduling model there are no deviations from workdays compared to manual scheduling there are deviations of 8 working days with excess 6 days and lack of 2 working days can be seen in Table 2. The new schedule considers the holidays, night shifts and workday needs of each nurse, so the new schedule has paid attention to the nurses' fatigue factor.
Table 3. Comparison of Ward Room Worker Day Deviations.
Manual Schedule Class Nursing Room
Nurse Mor
ning
Aftern
oon
Nig
ht
Total KB KK
1 15 8 0 22 1 0
2 11 6 6 23 1 0
3 9 6 7 22 0 0
4 6 7 9 22 0 0
5 8 7 7 22 0 0
6 7 7 8 22 0 0
7 8 5 9 22 0 0
8 9 6 7 22 0 0
9 9 8 6 22 0 0
10 8 9 5 22 0 0
11 11 4 7 22 0 0
12 7 6 8 21 0 1
13 4 11 6 21 0 1
14 6 9 7 22 0 0
Amount 6 2
Total Deviations 8
New Model Class Nursing Room
Nurse Mor
ning
Aftern
oon
Nig
ht
Total KB KK
1 8 9 5 22 0 0
2 8 7 7 22 0 0
3 8 7 7 22 0 0
4 10 7 5 22 0 0
5 15 2 5 22 0 0
6 8 7 7 22 0 0
7 8 7 7 22 0 0
8 9 6 7 22 0 0
9 4 10 8 22 0 0
10 7 11 4 22 0 0
11 9 6 7 22 0 0
12 12 4 6 22 0 0
13 9 6 7 22 0 0
14 6 7 9 22 0 0
Amount 0 0
Total Deviations 0
Manual Schedule Ward Room
Nurse Mor
ning
Aftern
oon
Nig
ht
Total KB KK
1 20 3 0 23 1 0
2 10 6 7 23 1 0
3 7 6 8 21 0 1
4 5 9 8 22 0 0
5 6 8 9 23 1 0
6 8 7 9 24 2 0
7 5 7 9 21 0 1
8 9 5 8 22 0 0
9 5 7 10 22 0 0
10 5 7 9 21 0 1
11 7 6 8 21 0 1
12 8 5 9 22 1 0
13 5 8 8 21 0 1
14 6 7 9 22 0 0
Amount 9 5
Total Deviations 14
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10528
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative
Commons Attribution-NonCommercial 4.0 International License. 128
In table 3. it can be seen that a comparison of scheduling models between manual models and the new scheduling model in September shows that the new scheduling model provides a more equal number of workdays to all nurses. In the new scheduling model there are no deviations from workdays compared to manual scheduling there are deviations of 13 workdays with excess 9 days and a deficiency of 5 workdays can be seen in Table 3. The new schedule considers the holidays, night shifts and workday needs of each nurse so that the new schedule has paid attention to the fatigue factor of nurses.
IV. CONCLUSION AND SUGGESTION
1. Conclusion
Based on the results, the conclusions that can be drawn are:
Nurse scheduling modeling can solve nurses scheduling problems that minimize the deviation (deviation) of each nurse's workday.
Development of the scheduling model with the rules recommended by the General Hospital in modeling nurses scheduling more effectively. This can be seen in the new schedule for the number of nurses fulfilled, setting the number
of night shifts and the need for days off accordingly.
The scheduling model that produces a new schedule already regulates the need for days off and night shifts so that attention is paid to the nurse's fatigue.
2. Suggestion
This research can be developed to solve scheduling problems with a greater number of nurses.
Research can also be developed by completing various cases of nurse scheduling by considering nurse requests on night shifts and days off so that the level of nurse satisfaction is met.
V. REFERENCES
Angeline, A., Iryanto, I., & Tarigan, G. (2014).
Penerapan metode branch and bound dalam
menentukan jumlah produksi optimum pada
cv. Xyz. Saintia Matematika , 2 (2), 137-
145.
Bakhtiar, H. T., & Jaharuddin, J. (2017). The nurse
scheduling problem: a goal programming
and nonlinear optimization approaches. IOP
Conference Series: Materials Science and
Engineering .
Hakim, L., Bakhtiar, T., & Jaharuddin, J. (2017). The
nurse scheduling problem: a goal
programming and nonlinear optimization
approaches. Materials Science and
Engineering , 1-7.
Hasan, M. M., & Arefin, M. R. (2017). Aplication Of
Linear in Scheduling Problem. Dhaka Univ.
J. Sci , 145-150.
Jafari, H., & Salmasi, N. (2015). Maximizing the
nurses’ preferences in nurse scheduling
problem: mathematical modeling and a
meta-heuristic algorithm. J Ind Eng Int .
KOÇ, B. A., & AKTAN, M. (2019). The Solution of
Nurse Scheduling Problem with Simulated
Annealing Algorithm. Journal of Scientific
and Engineering Research , 6 (4), 153-160.
Nur, W., & Abdal, N. M. (2016). Penggunaan Metode
Branch and Bound dan Gomory Cut.
JURNAL SAINTIFIK , 2 (1), 9-15.
Rafeek, F. S., & Siswanto, N. (2015). Solving Course
Timetable Problem by using Integer Linear
Programming (Case Study IE Department of
New Model Ward Room
Nurse Mor
ning
Aftern
oon
Nig
ht
Total KB KK
1 10 7 5 22 0 0
2 13 4 5 22 0 0
3 14 2 6 22 0 0
4 10 6 6 22 0 0
5 12 3 7 22 0 0
6 9 5 8 22 0 0
7 8 7 7 22 0 0
8 11 5 6 22 0 0
9 13 3 6 22 0 0
10 11 5 6 22 0 0
11 8 6 8 22 0 0
12 9 5 8 22 0 0
13 11 4 7 22 0 0
14 10 5 7 22 0 0
Amount 0 0
Total Deviations 0
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10528
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative
Commons Attribution-NonCommercial 4.0 International License. 129
ITS). International Seminar on Science and
Technology .
Sumathi, P. (2016). A new approach to solve linear
programming problem with intercept values.
Journal of Information & Optimization
Sciences .
Suryawan, G., Tastrawati, N. K., & Sari, K. (2016).
Penerapan branch and bound algorithm
dalam optimalisasi produksi roti. E-Jurnal
Matematika , 5 (4), 148-155.
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10512
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative
Commons Attribution-NonCommercial 4.0 International License. 130
Evaluation of the Satisfaction of Users of
Weather Forecast Systems with the Service
Quality Method
Eva Rianti
Universitas Putra Indonesia YPTK Padang
Padang, Indonesia
Syafrika Deni Rizky
Universitas Putra Indonesia YPTK Padang
Padang, Indonesia
Fariz Haris Nugraha
Universitas Putra Indonesia YPTK Padang
Padang, Indonesia
Submitted: Feb 29, 2020
Accepted: Apr 8, 2020
Published: Apr 8, 2020
Abstract—This research was conducted to test the user's satisfaction of the weather
forecast system that had been previously applied to the Teluk Bayur BMKG. The
weather forecasting service at the Padang Bayur Padang BMKG has basically fulfilled
the objectives of the company. However, over time, evaluation is needed so that the
effectiveness of the use of the system can be known, so the system can continue to be
developed. To overcome these problems, it is necessary to evaluate the performance of
the system. One way is to measure user satisfaction by using Service Quality. This
method is one of the techniques used to measure the level of user satisfaction with the
system. Based on the results of the measurement of the quality of the weather forecast
system, the results are obtained that the system user is satisfied with the system
performance, but there are several indicators on the system whose performance is not
fully maximized, so that there needs to be improvements and improvements of the
system performance.
Keywords—Service Quality, User Satisfaction, System, Evaluation
I. INTRODUCTION
BMKG Teluk Bayur Padang as one of the
companies that has implemented many information
systems to help the work process. In the Teluk
Bayur Padang BMKG there is an information
system that can help manage weather data. The
information system is called the Weather Forecast
Information System. BMKG Teluk Bayur Padang
realizes that information about weather data is very
important because this information is very useful
for parties involved in the decision making process,
especially for shipping traffic, crossings, and so on.
The system is used to help input weather data,
which results will be displayed on the BMKG
system.
The system has experienced many developments in
terms of features and functions that will facilitate
the work of its users (Nugroho, 2019). One way to
optimize and improve the performance of the
system is by knowing the satisfaction from the
user's side. System performance can be seen from
user satisfaction in using the information system.
Therefore, in this study an evaluation of the user
satisfaction of the Weather Forecast Information
System can be seen in the system's performance.
One method used to assess user satisfaction is the
service quality method. This service quality method
is one of the methods used to measure the level of
user satisfaction with information systems.
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10512
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative
Commons Attribution-NonCommercial 4.0 International License. 131
II. LITERATURE REVIEW
2.1 BASIC CONCEPTS OF INFORMATION SYSTEMS
A system can be interpreted as a collection or set of
elements, components, or variables that are
organized, interact with each other, interdependent
with each other (Rianti, 2016). System theory says
that every element of an organization is important
and must receive full attention so that managers can
act more effectively (Zefriyenni, 2020)
2.1.2 Information System
Information system is a collection of software and
hardware devices and human devices that will
process data using the hardware and software
(Yenila, 2019)
2.2 Service Quality
Service Quality method is a method of measuring
the quality of service most often used (Miranda,
2018). Assessing the quality of service of a service
provider based on five quality dimensions that are
often called Q-RATER the q-rater is:
1. Reliability, is the ability to provide the
promised service reliably and accurately.
2. Assurance namely guarantees and
trustworthiness owned by the company,
including the knowledge and courtesy of
employees in serving consumers, as well as
their ability to maintain consumer confidence.
3. Tangible (physical evidence), the appearance
of facilities and physical equipment used to
provide or provide services, including the
appearance of physical facilities, equipment,
workers or communication tools.
4. Emphaty that individual attention provided by
the company to consumers, including the ease
of making good relationships, personal
attention and understanding customer needs.
5. Responsiveness, namely the desire to help
consumers and provide services quickly,
namely the desire of staff to help customers
and provide services with responsiveness.
2.2.1 Service Quality Measurement
Quality measurements in the service quality model
are based on a multi-item scale designed to measure
expectations on five dimensions for service quality
(Ardhyani, 2017). Service quality evaluation using
the service quality model includes calculation of the
difference between the values given by customers,
for each pair of questions related to expectations
and perceptions (R Govindaraju, 2016). Service
quality scores for each pair of questions related to
expectations and perceptions (Winarto, 2017).To
measure the score of service satisfaction level the
following formula will be used:
Service Quality Score=Perception Score-Expected
Score
KL = P – E
KL: Service quality score
P: Customer perception score
E: Customer expectation score
Data obtained through service quality instruments
can be used to calculate service quality gap scores
at various levels in detail (Krisdayanti, 2017):
a. Item - by item analysis, for example P1 - E1, P2
- E2, and so on.
b. Dimension - by dimension analysis, for example
(P1 + P2 + P3 + P4 / 4) - (E1 - E2 - E3-E4 / 4) where
P1
through P4 and E1 to E5 reflect empathy
questions of perception and expectations related
to certain dimensions.
c. Single size calculation for servqual service
quality, that is (P1 + P2 + + P22 / 22 - (E1 + E2 +
..... + E22 / 22).
2.2.2 Steps for calculating the Sevice Quality
method
The steps that must be taken in this Service quality
method are as follows (Kuspriyono, 2017):
1. Determine the list of service attributes to be
measured. To determine what attributes will
be displayed, service providers can start by
referring to the five main dimensions of
service quality as research variables.
Attributes made in the form of questions in
accordance with the intent of the variables of
each study.
2. Knowing the customer's opinion about these
attributes. Respondents were asked to answer
the question of each attribute in the list,
namely how important or how much customer
expectations of these attributes, how much
weight is given for each service provided and
how well the service performance felt by
customers after using it.
3. For each customer, determine the Servqual
score for each question / attribute with the
following equation:
Si =Pi - Ei; i =1,2,3,...,n
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10512
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative
Commons Attribution-NonCommercial 4.0 International License. 132
Pi: Perceptual value given by the
customer to the i question
Ei: Value of the expectation given by the
customer for question I
N: Question number
4. For customers, add up the Servqual value
obtained for each dimension, then divide the
number by the number of questions or
attributes on that dimension
5. For each customer, multiply the Ski value by
the weight (wi) given for each dimension to
get the weighted Servqual value (Sqi) for each
dimension.
6. TSQ calculations for all customers are added
up and divided by n to get the average
Servqual value.
III. PROPOSED METHOD
In this study, the research framework can be seen
from Figure 2,
Field survey
Identification of problems
Study Literature
Service Quality Analysis
Instrument Design
Service Quality Test Information
System
Analyze test results
Recommendation
Figure 2.Service Quality Method Chart
Quotioner data analysis is performed by means of
quantitative descriptive which is a direct
explanation of the results of calculations using
formulas on service quality.
IV. RESULT AND DISCUSSION
4.1. System Analysis
System analysis is carried out aiming to find out the
problems that occur in library information systems,
as a basis for being able to measure the quality of a
system from theories that have been studied (Dewi
S Jannah, 2018)
.
4.2. System Quality Measurement Using Service
quality
Service quality measurement data encompasses
user expectations and perceptions of the
performance of the Weather Forecast System at the
Padang Bayur Padang BMKG. Assessments are
grouped into 5 Iikert scales as follows:
a. For expectations:
1. Value 1: Very unimportant.
2. Value 2: Not important
3. Value 3: Neutral
4. Value 4: Penting
5. Value 5: Very Important
b. For expectations:
1. Value 1: Very unimportant.
2. Value 2: Not important
3. Value 3: Neutral
4. Value 4: Important.
5. Value 5: Very important
3.3.Questionnaire Results
In this study, there were 6 respondents, consisting
of 6 employees in the Teluk Bayur Padang BMKG
office who used the Weather Forecast System. The
results of the questionnaire can be seen as follows:
A Expectation Value
The results of the questionnaire for the expected
value can be seen in the following table
Table 1 .Expectation Questionnaire Results
No
Qual
ity
Dim
ensi
ons
Indicator SP P N TP STP
1
Rel
iabil
ity
Website can
process data
well
1 3 2 0 0
2
The features
available on
the website
can work well
3 2 1 0 0
3
Attractive
table features
and designs
2 2 2 0 0
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10512
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative
Commons Attribution-NonCommercial 4.0 International License. 133
4
The language
used is
consistent on
every page
1 4 0 1 0
5
Information
from this
website is
accurate and
error free
1 3 2 0 0
6
Res
ponsi
ven
ess
Text from
every page can
be responded
to and read
clearly
0 4 1 1 0
7
The choice of
buttons and
menus on the
website can
respond well
0 4 2 0 0
8
The login
process can be
done quickly
1 4 1 0 0
9
Ass
ura
nce
Website access
rights can be
guaranteed and
cannot be used
by parties who
are not entitled
to use
1 2 3 0 0
10
The process of
inputting data
can be done
safely and
smoothly
1 2 2 1 0
11
Em
phat
y
The operator
can see the
time and date
of use of the
website
0 4 2 0 0
12
Users can be
interpreted by
operators
0 4 2 0 0
13
Tan
gib
le
System hint service for
new users
0 3 3 0 0
amount 11 41 23 3 0
B Perception Value
The results of the questionnaire for the value of
perception can be seen in the following table:
Table 2 .Perception Value Questionnaire Results
N
o
Dimention Of
Quality Indicator
S
P P N
T
P
ST
P
1
Reli
abil
ity
Website
can do
data
processin
g well
4 1 1 0 0
2
The
features
available
on the
website
can work
well
5 0 0 1 0
3
Attractiv
e table
features
and
designs
4 1 1 0 0
4
the
language
used is
consisten
t on
every
page
2 3 0 1 0
5
Informati
on from
this
website
is
accurate
and error
free
3 1 2 0 0
6
Resp
on
siveness
Text
from
every
page can
be
responde
d to and
read
clearly
1 5 0 0 0
7
The
choice of
buttons
and
menus on
the
website
can
respond
well
2 3 1 0 0
8
The login
process
can be
done
quickly
2 3 0 1 0
9
Ass
ura
nce
Website
access
rights
can be
guarante
ed and
cannot be
used by
parties
who are
not
entitled
to use
4 1 1 0 0
10
The
process
of
inputting
data can
be done
safely
and
smoothly
3 1 1 1 0
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10512
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative
Commons Attribution-NonCommercial 4.0 International License. 134
11
Em
ph
aty
Operator
s can see
the time
and date
of
website
usage
1 4 1 0 0
12
Users
can be
interprete
d by
operators
1 4 1 0 0
13
Tan
gib
le System
hint
service
for new
users
2 2 2 0 0
Jumlah 3
4
2
9
1
1 4 0
C. Expectation Value
1. Calculates the total score of each indicator.
Carried out by: Total score= P5x5 + P4x4 + P3x3
+P2x2+ P1x1. Where the results are as follows:
Indicator 1 = (5 x 4) + (4 x 1) + (3 x 1) +
(2 x 0) + (1 x 0) = 27
Indicator 2 = (5 x 5) + (4 x 0) + (3 x 0) +
(2 x 1) + (1 x 0) = 27
Indicator 3 = (5 x 4) + (4 x 2) + (3 x 2) +
(2 x 0) + (1 x 0) = 27
Indicator 4 = (5 x 2) + (4 x 3) + (3 x 0) +
(2 x 1) + (1 x 0) = 24
Indicator 5 = (5 x 3) + (4 x 1) + (3 x 2) +
(2 x 0) + (1 x 0) = 25
Indicator 6 = (5 x 1) + (4 x 5) + (3 x 0) +
(2 x 0) + (1 x 0) = 25
Indicator 7 = (5 x 2) + (4 x 3) + (3 x 1) +
(2 x 0) + (1 x 0) = 25
Indicator 8 = (5 x 2) + (4 x 3) + (3 x 0) +
(2 x 1) + (1 x 0) = 24
Indicator 9 = (5 x 4) + (4 x 1) + (3 x 1) +
(2 x 0) + (1 x 0) = 27
Indicator 10 = (5 x 3) + (4 x 1) + (3 x 1) +
(2 x 1) + (1 x 0) = 24
Indicator 11 = (5 x 1) + (4 x 4) + (3 x 1) +
(2 x 0) + (1 x 0) = 24
Indicator 12 = (5 x 1) + (4 x 4) + (3 x 1) +
(2 x 0) + (1 x 0) = 24
Indicator 13 = (5 x 2) + (4 x 2) + (3 x 2)
+ (2 x 0) + (1 x 0) = 24
2. Divide the total score by the number of respondents
Done by: hope value = total score: number of
respondents. Expectation value indicates the value
of service expected by the user on the performance
of the Weather Forecast System at BMKG Teluk
Bayur Padang. The expected value of each indicator
will be presented in the following table:
Table 3. User Expectation Value
N
o Indicator
Expectation
5 4 3 2 1
Tota
l
Expec
tati
on V
alue
SP
P
N
TP
ST
P
1
Website
can do
data
processin
g well
4 1 1 0 0 27 4,5
0
2
The
features
available
on the
website
can work
well
5 0 0 1 0 27 4,5
0
3
Attractive
table
features
and
designs
4 1 1 0 0 27 4,5
0
4
the
language
used is
consistent
on every
page
2 3 0 1 0 24 4,0
0
5
Informati
on from
this
website is
accurate
and error
free
3 1 2 0 0 25 4,1
7
6
Text from
every
page can
be
responde
d to and
read
clearly
1 5 0 0 0 25 4,1
7
7
The
choice of
buttons
and
menus on
the
website
can
respond
well
2 3 1 0 0 25 4,1
7
8
The login
process
can be
done
quickly
2 3 0 1 0 24 4,0
0
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10512
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative
Commons Attribution-NonCommercial 4.0 International License. 135
9
Website
access
rights can
be
guarantee
d and
cannot be
used by
parties
who are
not
entitled to
use
4 1 1 0 0 27 4,5
0
10
The
process
of
inputting
data can
be done
safely
and
smoothly
3 1 1 1 0 24 4,0
0
11
Operators
can see
the time
and date
of
website
usage
1 4 1 0 0 24 4,0
0
12
Users can
be
interprete
d by
operators
1 4 1 0 0 24 4,0
0
13
System
hint
service
for new
users
2 2 2 0 0 24 4,0
0
Average
25,1
5
4,1
9
D.Perception Value
1. Calculate the total score of each indicator.
Done by means of a total score = P5X5 + P4x4 +
P3x3 + P2x2 + P1x1
Where the results are as follows:
Indicator 1 = (5 x 1) + (4 x 3) + (3 x 2) + (2 x 0) +
(1 x 0) = 23
Indicator 2 = (5 x 3) + (4 x 2) + (3 x 1) + (2 x 0) +
(1 x 0) = 26
Indicator 3 = (5 x 2) + (4 x 2) + (3 x 2) + (2 x 0) +
(1 x 0) = 24
Indicator 4 = (5 x 1) + (4 x 4) + (3 x 0) + (2 x 1) +
(1 x 0) = 23
Indicator 5 = (5 x 1) + (4 x 3) + (3 x 2) + (2 x 0) +
(1 x 0) = 23
Indicator 6 = (5 x 0) + (4 x 4) + (3 x 1) + (2 x 1) +
(1 x 0) = 21
Indicator 7 = (5 x 0) + (4 x 4) + (3 x 2) + (2 x 0) +
(1 x 0) = 22
Indicator 8 = (5 x 1) + (4 x 4) + (3 x 1) + (2 x 0) +
(1 x 0) = 24
Indicator 9 = (5 x 1) + (4 x 2) + (3 x 3) + (2 x 0) +
(1 x 0) = 22
Indicator 10 = (5 x 1) + (4 x 2) + (3 x 2) + (2 x 1)
+ (1 x 0) = 21
Indicator 11 = (5 x 0) + (4 x 4) + (3 x 2) + (2 x 0)
+ (1 x 0) = 22
Indicator 12 = (5 x 0) + (4 x 4) + (3 x 2) + (2 x 0)
+ (1 x 0) = 22
Indicator 13 = (5 x 1) + (4 x 3) + (3 x 3) + (2 x 0)
+ (1 x 0) = 21
2. Divide the total score by the number of
respondents
Done with the formula: expectation value = total
score: number of respondents
Table 4. User Perception Value
No Indikator
Persepsi
5 4 3 2 1
To
tal
Perc
epti
on
Valu
e
SP
P N TP
ST
P
1
Website can
do data
processing
well
1 3 2 0 0 23 3,83
2
The features
available on
the web can
work well
3 2 1 0 0 26 4,33
3
Attractive
table
features and
designs
2 2 2 0 0 24 4,00
4
the
language
used is
consistent
on every
page
1 4 0 1 0 23 3,83
5
Information
from this
website is
accurate and
error free
1 3 2 0 0 23 3,83
6
Text from
every page
can be
responded
to and read
clearly
0 4 1 1 0 21 3,50
7
The choice
of buttons
and menus
on the
website can
respond
well
0 4 2 0 0 22 3,67
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10512
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative
Commons Attribution-NonCommercial 4.0 International License. 136
8
The login
process can
be done
quickly
1 4 1 0 0 24 4,00
9
Website
access
rights can
be
guaranteed
and cannot
be used by
parties who
are not
entitled to
use
1 2 3 0 0 22 3,67
10
The process
of inputting
data can be
done safely
and
smoothly
1 2 2 1 0 21 3,50
11
Operators
can see the
time and
date of
website
usage
0 4 2 0 0 22 3,67
12
Users can
be
interpreted
by operators
0 4 2 0 0 22 3,67
13
System hint
service for
new users
0 3 3 0 0 21 3,50
Average 22,62 3,77
3. Calculate Service Quality Score
Servqual Score (Gap Score) = Perception Value
- Expectation Value
Where the results are as follows:
Gap Score 1 = 3,83 – 4,50 = - 0,67
Gap Score 2 = 4,33 – 4,50 = - 0,17
Gap Score 3 = 4,00 – 4,50 = - 0,50
Gap Score 4 = 3,83 – 4,00 = - 0,17
Gap Score 5 = 3,83 – 4,17 = - 0,34
Gap Score 6 = 3,50 – 4,17 = - 0,67
Gap Score 7 = 3,67 – 4,17 = - 0,50
Gap Score 8 = 4,00 – 4,00 = 0,00
Gap Score 9 = 3,67 – 4,50 = - 0,83
Gap Score 10 = 3,50 – 4,00 = - 0,50
Gap Score 11 = 3,67 – 4,00 = - 0,33
GapSkor 12 = 3,67 – 4,00 = - 0,33
Gap Score 13 = 3,50 – 4,50 = - 0,50
Table.5 Service quality scores User Satisfaction
Rating
No Indicator
Per
cepti
on
Val
ue
Expec
tati
o
n V
alue
Gap Score
1
Website can do
data processing
well
3,83 4,50 -0,67
2
The features
available on the
website can work
well
4,33 4,50 -0,17
3
Attractive table
features and
designs
4,00 4,50 -0,50
4
the language used
is consistent on
every page 3,83 4,00 -0,17
5
Information from
this website is
accurate and error
free
3,83 4,17 -0,34
6
Text from every
page can be
responded to and
read clearly
3,50 4,17 -0,67
7
The choice of
buttons and menus
on the website can
respond well
3,67 4,17 -0,50
8
The login process
can be done
quickly
4,00 4,00 0,00
9
Website access
rights can be
guaranteed and
cannot be used by
parties who are not
entitled to use
3,67 4,50 -0,83
10
The process of
entering data can
be done safely and
smoothly
3,50 4,00 -0,50
11
Operators can see
the time and date
of website usage 3,67 4,00 -0,33
12
Users can be
interpreted by
operators
3,67 4,00 -0,33
13
System hint
service for new
users
3,50 4,00 -0,50
amount 49,00 54,51 -5,51
Gap Value Maximum Score 0,00
Gap Value Minimum Score -0,83
E. Cartesian diagram
In this study there are two variables represented by
the letters X and Y, where X is the level of user
perception of the performance provided by the
system, while Y is the level of user expectations of
the performance provided by the system.
Furthermore, the levels of these elements are
elaborated and divided into 4 parts into a Cartesian
diagram
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10512
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative
Commons Attribution-NonCommercial 4.0 International License. 137
The formula: Ẋ = ΣXi: K and Ẏ = ΣYi: K
Ẋ = average score of implementation level
Ẏ = average score of importance
K = number of factors / attributes
Where the results are as follows
1. For User Perception
Ẋ = 49.00: 6 = 3.77
2. For User Expectations
Ẏ = 54.51: 6 = 4.19
From the above calculation, it is obtained that the
value of X = average score of the perception level
and Y = the average score of the expectation level
presented in the following Table.6:
Table.6 Values Ẋ and Cart Cartesian diagrams
No Variabel Information score
1 Ẋ Implementation level
(user perception) 3,77
2 Ẏ Level of importance
(user expectations) 4,19
This diagram consists of four quadrants: the first
quadrant is located in the upper left, the second
quadrant in the upper right, the third quadrant in the
lower left, and the fourth quadrant in the lower
right. Each variable in the quadrant can be
explained as follows:
1. Quadrant I
This is an area that contains factors that are
considered important by the user but in reality these
factors are not in line with what the user expects
(the level of satisfaction obtained is still very low).
The variables included in this quadrant must be
increased.
2. Quadrant II
This is an area that contains factors that are
considered important by users and factors that are
considered by users to be in accordance with what
they feel so that the level of satisfaction is relatively
higher. The variables included in this quadrant must
be maintained because all of these variables make
the indicator superior to the user.
3. Quadrant III
This is an area where the factors that are considered
less important by the user and in fact the
performance is not too special. The increase in the
variables included in this quadrant can be
reconsidered because the effect on the benefits felt
by the user is very small.
4. Quadrant IV
This is an area that contains factors that are
considered less important by the user and are overly
excessive. The variables included in this quadrant
can be reduced.
Figure 1: Cartesian Diagram of Satisfaction
Assessment
Figure.2 Matrix of Expectations and Perceptions
F. Distribution and percentage of users' perceptions
of service
Table.7 Distribution and Percentage of User
Satisfaction
Satisfaction
Response
Distributio
n
Percentag
e (%)
5 = Very satisfied 11 14,1
4 = Satisfied 41 52,6
3 = Neutral 23 29,5
2 = Not satisfied 3 3,8
1 = Very
Dissatisfied 0 0,0
Total 78 100
G.Quality Performance of Service Attributes
From the data obtained that the average servqual
score is -0.42 meaning that the gap score between
user expectations and what users feel is negative.
This means that the average new user expectations
are met by the Weather Forecast System with a
score of 0.42.
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10512
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative
Commons Attribution-NonCommercial 4.0 International License. 138
The gap score of each indicator and the average gap
score of the user satisfaction assessment by the
Weather Forecast System will be shown in the
following Table.8:
No Indicator Gap Score Gap
Average
1 Website can do data
processing well -0,67
-0,42
2 The features available on
the website can work well -0,17
3 Attractive table features
and designs -0,50
4 the language used is
consistent on every page -0,17
5
Information from this
website is accurate and
error free
-0,34
6
Text from every page can
be responded to and read
clearly -0,67
7
The choice of buttons and
menus on the website can
respond well
-0,50
8 The login process can be
done quickly 0,00
9
Website access rights can
be guaranteed and cannot
be used by parties who are
not entitled to use
-0,83
10
The process of inputting
data can be done safely and
smoothly -0,50
11 Operators can see the time
and date of website usage -0,33
12 Users can be interpreted by
operators -0,33
13 System hint service for
new users -0,50
H. Cartesian diagram analysis
1. Quadrant I (Top Priority)
The indicators that are in this quadrant are
considered very important by the user but the
system performance is not satisfactory. Indicators
in this quadrant become the main priority for
immediate improvement and improved
performance. The indicators in quadrant 1 are as
follows:
a. Indicator 9: Website access rights can be
guaranteed and cannot be used by parties who are
not entitled to use
2. Quadrant II (Maintain Performance)
The indicators in this quadrant are considered very
important by the user and the performance of the
system is very satisfying. The indicators included in
this quadrant must be maintained because the
performance indicators are considered superior to
users.
a. Indicator 1: Website can do data processing well
b. Indicator 2: The features on the website can work
well
c. Indicator 3: Attractive table designs and features
3. Quadrant III (Low Priority)
The indicators in this quadrant are considered not
too important for the user and the performance of
the system is not satisfactory. The increase of the
indicators in this quadrant can be reconsidered
because the effect on the benefits felt by the user is
very small. The indicators in quadrant III are as
follows:
a. Indicator 6: The writing of the text on each page
can be responded to and read clearly
b. Indicator 7: The choice of buttons and menus on
the website can respond well
c. Indicator 10: The process of inputting data can be
done safely and smoothly
d. Indicator 11: Operators can see the time and date
of website usage
e. Indicator 12: Users can be interpreted by
operators
f. Indicator 13: Service hints on the system for new
users
4. Quadrant IV (Excessive)
The indicators in this quadrant are considered not
too important by the user but the system is
satisfactory. Performance indicators in this
quadrant may be reduced because the user considers
not too important. The indicators in this quadrant
are as follows:
a. Indicator 4: The language used is consistent on
every page
b. Indicator 5: Information from this website is
accurate and error free
c. Indicator 8: Login process can be done quickly
V. CONCLUSIONAND SUGGESTION
A. Conclusion
Processing data taken from the results of the
questionnaire found that 52.6% of system users
feel satisfied with the performance of the
system and 14.1% of users rate very satisfied,
this shows that generally system users feel
satisfied with the current system performance.
Based on the evaluation that has been done, the
following conclusions are obtained:
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10512
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative
Commons Attribution-NonCommercial 4.0 International License. 139
1. From the Cartesian diagram image that
has been made before then the results are
found:
1. In quadrant I, contain indicator
question 9
2. In quadrant II, contain indicator
questions 1, 2, and 3
3. In quadrant III, contain indicators
questions 6, 7, 10, 11, 12, and 13
4. In quadrant IV, contain indicators
questions 4, 5, and 8
2. From the evaluation results of the
Cartesian Diagram we can find out that
the indicator that is in quadrant 1 is the
indicator question 9, namely website
access rights can be guaranteed and
cannot be used by unauthorized parties, a
major priority in improving system
performance, due to the performance of
the indicator this is considered important
by the user of the system but in reality this
indicator is not in line with what is
expected by the user, as for othindicators
that need to be considered to improve
performance is the indicator that is in the
third quadrant, because the benefits felt by
the user to the system feels very small, As
for the indicators are:
a. Indicator 6: The writing on each
page is responded to and read
clearly
b. Indicator 7: Choice of buttons and
menus on the website can respond
well
c. Indicator 10: The data input process
is carried out safely and smoothly
d. Indicator 11: Operators can see the
time when using the website
e. Indicator 12: Users can be
interpreted by operators
f. Indicator 13: Service hints on the
system for new users
3. Based on the results of measuring the
quality of the system using the Service
Quality method it can be concluded
that this method is quite effective to be
used in the calculation of the system
quality assessment, because a
company can know the level of system
performance and the level of needs of
users so that system performance can
be improved even better.
B. Suggestion
Based on the results of research conducted,
suggestions that can be given include:
1. It is expected that the IT team at the
company will be able to improve and
improve the performance of the system
based on the results of the evaluation
given previously, especially on the
indicators in quadrants 1 and 3.
2. Improving system performance can be
done by improving the quality of service
on each indicator and adding features that
it feels need to be added in order to
improve system performance.
3. Evaluation of user satisfaction is expected
to be carried out regularly and
continuously so that it can be known what
indicators of service quality need to be
improved, because the hopes and
perceptions of users are increasingly
developing over time.
VI. REFERENCES
ARDHYANI, I. (2017). PENGUKURAN KUALITAS
LAYANAN DENGAN HIGHER EDUCATION
SERVICE QUALITY (HIEDQUAL). TEKNIKA:
ENGINEERING AND SAINS JOURNAL , 25-32.
Dewi S Jannah, J. (2018). Analisis dan Perancangan
Sistem Informasi Manajemen Aset Tetap Pada
PT. Metis Teknologi Corporindo. JUST IT:
Jurnal Sistem Informasi, Teknologi Informasi
dan Komputer , 81-91.
Hashem, H. F. (2009). Adaptive technique for
human face detection using HSV color space
and neural networks. National Radio Science
Conference. IEEE.
Krisdayanti. (2017). Pengaruh Kualitas Layanan
Dan Kepuasan Konsumen Terhadap Loyalitas
Konsumen Pada Minimarket Kertapati Jaya
Indah Palembang.[Skripsi]. Doctoral
Dissertation UIN Raden Fatah Palembang .
Kuspriyono. (2017). Pengaruh Kualitas Informasi
Web dan Kualitas Layanan Online Terhadap
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10512
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative
Commons Attribution-NonCommercial 4.0 International License. 140
Citra Bukalapak. com. Junal Perspektif , 56-
62.
Miranda, S. (2018). Perceived Service Quality and
Customer Satisfaction : A Fuzzy set QCA
Approach in the Railway sector. Journal Of
Bussiness Research , 371-377.
Nugroho, S. (2019). Analisis Iklim Ekstrim Untuk
Deteksi Perubahan Iklim di Sumatera Barat.
Jurnal Ilmu Lingkungan , 7-14.
Pittsburgh, U. o. (2019, February 18). University of
Library System. Retrieved from Citation
Styles: APA, MLA, Chicago, Turabian, IEEE:
APA 6th Edition:
https://pitt.libguides.com/c.php?g=12108&p=
64730
R Govindaraju, W. (2016). Pengembangan Model
Evaluasi Kualitas Layanan Sistem E-
Government. Journal of Technology
Management , 196-205.
Rianti, E. (2016). Support, Designing Aplication
Decision Support System for Science
Olympics Participant Election In SMPN
Sijunjung Using Analitical Hierarky Proses.
Jurnal Sains dan Informatika : Research Of
Science and Informatic , 22-26.
Winarto, W. (2017). Persepsi Kualitas Layanan
Warung Internet di Kota Medan. Jurnal
Ilmiah METHONOMI , 107-115.
Yenila, F. (2019). Analisa Sistem Informasi
Pemeliharaan Prasarana Jalan dan Jembaran
Dinas Pekerjaan Umum. Jurnal Teknologi ,
21-25.
Zefriyenni. (2020). Reclamation Tax Calculating at
Revenue Service, Regional Financial asset
Management (DPPKAD) Pariaman City.
Jurnal KomtekInfo , 41-46.
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10536
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative
Commons Attribution-NonCommercial 4.0 International License. 141
Analysis of Cadets Satisfaction to Medan
Aviation Polytechnic Service Using Quality
Function Deployment (QFD) Method Liber Tommy Hutabarat
POLTEKBANG MEDAN
Medan, Indonesia
Lisda Juliana Pangaribuan
AMIK MBP
Medan, Indonesia
Submitted: Mar 26, 2020
Accepted: Apr 10, 2020
Published: Apr 10, 2020
Abstract— Education is an important factor in the development of science and the
technology. Quality of service is an important element in the institution of education for
the quality service of education that will produce high-quality graduates as well. The
importance of service quality makes it a part of the Higher Education Quality Standards.
Cadets Satisfaction to services is a benchmark to find out the quality of service at Medan
Aviation Polytechnic. This research aims to determine the level of Cadets satisfaction to
Medan Aviation Polytechnic service towards learning management standard services
and standards for infrastructure facilities. The results of the study will be used as
references in developing quality standards of Medan Aviation Polytechnic. The method
used was Quality Function Deployment (QFD) method, with a total sample of 44 people.
Analysis of the data is done with test validity, reliability, planning matrices and matrix
House of Quality (HOQ). Analysis results shows the level of cadets satisfaction highest
is the dimension Realibility to value the satisfaction of 4.01 followed by the dimensions
of Assurance with the value of the satisfaction of 3.98, then the dimension Empaty to
value the satisfaction of 3.97 and then at Responsivness dimension with satisfaction
value 3.95 and the last is Tangibility dimension with satisfaction value 3.92.
Keywords— QFD; Cadets Satisfaction; PlanningMatrix; HoQ Matrix
I. INTRODUCTION
Education world’s is an important factor in the development of science, therefore technological progress must increasingly encourage educational institutions to provide good services for users of educational services in order to produce quality individuals. It is undeniable cadets satisfaction to education services should be kept. Therefore the level of quality of the education service system must always be improved.
Medan Aviation Polytechnic is the institution has the basic principle of meeting the demands and cadets needs because cadets needs can increase the productivity of education. Enhancement of Education productivity in Medan Aviation Polytechnic become main benchmarks for goals to be achieved, so analysis of Medan Aviation Polytechnic quality service satisfaction should be performed to obtain an
understanding of cadets satisfaction to infrastructure and process learningservices. The way that can be taken is by trying to know and understand the expectations, perceptions of cadets and characteristics of educational services so that the priorities of the service can be known. National Higher Education Standards become a reference in preparation quality standards for the learning process and infrastructure. Assessment of satisfaction for service quality is very important to be carried out in the preparation of standards for the learning process and infrastructure in the Quality Assurance System. In addition, the implementation of the standard learning processes and infrastructures is important part of the instrument of accreditation of study programs and institutional accreditation. The results of the assessment of satisfaction for the standards of the learning process and infrastructure can also be
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10536
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative
Commons Attribution-NonCommercial 4.0 International License. 142
input for institutions in order to improve and develop education.
Based on the Regulation of the Minister of Research, Technology, and Higher Education Republic Indonesia No. 44 of 2015 concerning National Standards of Higher Education, to get a goal of national education, efforts to improve the quality, amount, type, and specifications based on the use of facilities are in accordance with the characteristics of methods and forms of learning, must guarantee the implementation of the learning process and academic administrative services. Quality of service is a very important factor and has become an issue today. From this aspect it is known that the government plays an important role in encouraging national education through a variety of applicable rules and regulations.
Quality Function Deployment (QFD) or the distribution of quality functions is a tool used to support the implementation of TQM (Total Quality Management) and quality improvement programs to improve the performance of an educational institution that affects the improvement of consumer satisfaction or vice versa. This method is a method, a structured tool for assessing satisfaction so as to improve service quality is the Quality Function Deployment (QFD) method.
II. LITERATURE REVIEW
In research conducted by Kusumawardhani said that to improve service in institutional workshops is expected to be able to develop workshop services by paying attention to customer satisfaction (cadets) then services also need to be improved by improving and improving the quality of human resources such as training to support the ability of workshop assistants to handle problems in PUM workshops, however this research is limited to Workshop services so that input cannot be made to improve the quality standards of institutions. (Kusumawardhani & Shafiq, 2018).
According to Claudia, the proposed improvements can be given to increase customer satisfaction including checking the availability of goods in goods on a scheduled basis (5 hours) from the opening hours of the store, then the company's strategy in selecting suppliers who become suppliers for its outlets must be more complete and quality, scheduling checking of suggestion boxes or customer service. But the results of this study cannot be applied to other agencies because the data processed is only based on 1 type of respondent, namely consumers who have shopped at the SUPER INDO Tembalang hypermarket (Claudia Gita & Susanti, 2017).
In a study conducted by Andre Audi Havid et al. (2016) stated the service attribute "completeness of
facilities and infrastructure in information media (internet)" shows the broadest tolerance zone and the attribute "comfortable seating conditions" is the narrowest tolerance zone. The QFD results show that the priority of the technical response given is providing employee performance training. (Andre A et al., 2016)
Elements academic services include elements of academic staff such as lecturers, elements of academic support staff are laboratory staff and academic administrative staff. Of course, in addition to human resources, the availability of other resources that support academic activities in the form of facilities and infrastructure will largely determine the quality of academic services provided by Suhendar & Suroto (2014).
Service quality is a measure of how well the level of service provided is able to match customer expectations. Service quality is determined by the company's ability to meet the needs and desires of customers in accordance with customer expectations. (Tjiptono, 2012)
Various studies have been conducted on several types of services, and managed to identify five dimensions of characteristics used by customers in evaluating service quality. The five dimensions of service quality characteristics are Tangibles, Reliability, Responsiveness, Assurance and Empaty (Yamit, 2013)
Higher Education Standards used is standard on Minister of Research, Technology and Higher Education Regulation Republik Indonesia No. 44 of 2015 about Standar Nasional Pendidikan Tinggi (Republik Indonesia., 2015) and Amendment Minister of Research, Technology and Higher Education Republic Indonesia No. 50 of 2018. (Republik Indonesia, 2018)
According to Meriastuti (Ginting & Halim, 2012) Quality Function Deployment (QFD) is a way to improve the quality of goods or services by understanding the needs of consumers and then linking with technical provisions to produce goods or services produced. QFD is a practice for designing a process in response to customer needs. QFD allows organizations to prioritize customer needs, find innovative responses to those needs, and improve processes to achieve maximum effectiveness (Tutuhatunewa, 2010).
The House of Quality (HOQ) matrix is tools for QFD representation (Azizah N. I. et al., 2018). HOQ use matrix to connect Voice of Customer and technical response. (Sutawijaya & Pista, 2018) This matrix consists of two main parts namely the
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10536
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative
Commons Attribution-NonCommercial 4.0 International License. 143
horizontal and vertical parts. The Horizintal section of this matrix contains information relating to consumers and this is called the customer table. The vertical part of the matrix contains technical information in response to consumer input, and is called a technical table. Customer information about consumers to provide information in the formation of the QFD method, while information techniques are respondents needed from consumers who are useful for distributors.
The process in QFD is carried out by compiling one or more matrices called The House of Quality. This matrix explains what the customer's needs and expectations are and how to fulfill them. The matrix called House of Quality in general can be seen in Figure 1.
Figure 1. The House of Quality
(Source: Cohen, 2007) Population is a generalization area consisting of
objects and subjects that have certain qualities and characteristics determined by researchers to be studied and then drawn conclusions (Rusiadi et al., 2014).
Champion (1981) says that most statistical tests always include sample size recommendations. In other words, existing statistical tests will be very effective if applied to samples from 30 to 60. Even if the sample is above 500, it is not recommended to apply a statistical test. (Basic Statistics for Social Research, Second Edition)
According to Gay and Diehl (1992), for descriptive research, the sample is 10% of the population, correlational research, at least 30 population elements, causal comparative research, 30 elements per group, and for experimental research 15 elements per group. (Hasim & Yusuf, 2010)
Validity test is used to measure the validity or validity of a questionnaire, (Sugiono, 2010). Test the validity of this research using Microsoft Excel. Significance test is done by comparing the value of r arithmetic with rtable for degree of freedom.
df = n-2 (1)
Validity test
ProbabIlistik= TINV (α ; df) (2)
Rtable = 𝑃𝑟𝑜𝑏𝑎𝑏𝑖𝑙𝑖𝑠𝑡𝑖𝑘
√𝑑𝑓+ 𝑃𝑟𝑜𝑏𝑎𝑏𝑖𝑙𝑖𝑠𝑡𝑖𝑘2 (3)
The reliability test is used to measure the
consistency of respondents in answering the research
questionnaire. The questionnaire is said to be reliable
if the respondent's answer to the question is consistent
or stable over time. The reliability test is used to
determine the consistency of the measuring
instrument, whether the measuring device used is
reliable and remains consistent if the measurement is
repeated. Reliability can be determined by using the
following formula: Reabilitas = k / (k-1) * (1- Σσi2/ σt2 (.4)
The instrument reliability of this study was calculated by the Alpha Cronbach formula with the minimum requirement to be considered reliable is> 0.7. (Sugiono, 2010).
In the planning matrix calculations performed are: a. Measurement scale = total score of the assessment
of cadets' perceptions with a Likert scale which is modified as follows : (Abdulrahman F & Handayani, 2017) 1 = not very good / very unsatisfactory / never /
incomplete / not important 2 = not good / low / rare / less complete / less
important 3 = sufficient / sometimes / sufficiently complete /
important enough 4 = good / high / frequent / complete / important 5 = very good / very high / always / very complete
/ very important b. Level of importance = Scoring score that most
often appears c. The level of satisfaction of cadets was made to
determine the needs of cadets. Formula for calculate Voice of Cadets = Total score atribut / total sample.
Formula for calculating cadets satisfaction = Total score attribute / number of samples. d. Determination of Goal = Target Score desired e. Improvement ratios are made to determine the
magnitude of changes made by management. Repair ratio formula = Goal / Taruna satisfaction level.
f. Selling Point with conditions If the satisfaction level is> 3, the selling point is 1.5 If the satisfaction level is 2-3, the selling point is 1.2 If the satisfaction level is 0-2, selling point is 1 (5)
g. RowWeight is useful to find out the amount of improvement needs of cadets. Can be calculated by the formula: Row Weight = Importance * Repair Ratio * Point of Sale (6)
E
C
A D B
F
Technical Matrix
Technical Response
Technical Coleration
Relationship Planning MatrixCostummer need
and Benefit
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10536
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative
Commons Attribution-NonCommercial 4.0 International License. 144
h. Normalized Weight can be calculated by the formula: Normalized Weight = Row Weight for each attribute: Total Row Weight (7)
Technical Matrix is conducted to determine the strength of the relationship between technical responses and cadets needs. The correlation value can be seen from the amount of Contribution Contribution = (Numerical Value * Numerical Raw Weight) (8)
Each relationship has its own weight value, if the relationship is very strong then the weight value is 9, if the relationship is strong the weight value is 3, if the relationship is weak the weight value is 1, and if there is no relationship between the two, then the weight value is 0. (Kusumawardhani & Shafiq, 2018)
III. PROPOSED METHOD
This is a descriptive study using survey methods to describe the gap between the expected services with services that have been perceived and Quality Function Deployment method to describe the development of services that will be recommended according to the needs of cadets. In this study, data collection was obtained by searching literature, field research in the form of interviews, questionnaires or direct observation of the actual situation. The variables observed in this study are the quality of service implemented by Medan AVIATION POLYTECHNIC to meet the National Standards of Higher Education, especially in the National Education Standards. The stages can be seen in Figure 2.
Figure 2. Research Method
Algorithm of Quality Function Deployment
method (data analysis) :
A. The phase for collecting voice of customer 1. Knowing the attributes of the cadets desires 2. Knowing the perception of cadets 3. Conducting validity and reliability testing 4. Classifying attributes of cadets desires
B. Phase for compiling House of Quality Matrix
1. Make a Planning Matrix
2. Making Technical Responses as Characteristics of
Medan AVIATION POLYTECHNIC Services.
3. Determine the relationship of technical parameters
(how) and the desires of cadets (whats)
4. Calculate technical correlation (Technical matrix) and determine service priority
5. Making the House of Quality Matrix. .
Algorithm QFD (Quality Function Deployment) in this study can be seen in Figure 3.
Figure 3. Algorithm QFD
This recearch uses 3 (three) independent variables, namely: Students Perception (X1), Student expectation (X2), Technical characteristics of service (X3) and 1 (one) dependent variable which is a form of educational service quality (Y). 3 independent variable in this reseach are student perception (X1), students perception as a customer is defined that the evaluation of the form of educational services received begins even before it interacts with the service provider itself ; Student expectation (X2), Student expectation as a customer is defined as a form of educational services expected by a cadets (customer) that he feels deserves to be received if he uses these educational services; Technical characteristics of services (X3), Technical characteristics are defined as technical services provided by management to cadets.
The variables used in the study are based on the Regulation Minister of Research, Technology and Higher Education Republic Indonesia No. 44 of 2015 concerning National Standards of Higher Education which specifically only addresses the National Standards of Education. National Education Standards will be grouped according to the five dimensions of service quality described by Zethaml, et al (2013) are: Tangible, Reliability, Responsiveness, Assurance, Empaty.
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10536
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative
Commons Attribution-NonCommercial 4.0 International License. 145
Population in this recseach are all of cadets in Medan AVIATION POLYTECHNIC a number of 300 people. This research is a descriptive study, so the minimum number of samples is 10% of the population, so the number of samples used is 2 classes with a total of 44 people.
IV. RESULT AND DISCUSSION
The first step in the research is to look for the attributes desired by cadets of Medan AVIATION POLYTECHNIC for the learning management services and the infrastructure provided. From the results of interviews and observations obtained several questions used as a questionnaire as an attribute of the Taruna desires. The items for the questionnaire:
Learning Management Standards, namely: 1. Submission of information on requirements and needs of cadets, 2. Administrative Services, 3. Provision of learning equipment, 4. Distribution of Schedules, 5. The lecturer gives time to explain the Module / Teaching material, 6. The suitability of PBM implementation.
Infrastructure Standards: 7. Information regarding student services, 8. Provision of students equipment, 9. Cleanliness of the study room, 10. Condition conditioning in the study room, 11. Condition of Chairs and Tables in the study room (furniture), 12. Learning Infrastructure Functions such as LCD / Projector, 13. Layout of Tables and Chairs and other equipment, 14. Lighting the study room, 15. Cleanliness in the Dormitory and the Dormitory environment, 16. Condition of Air Conditioning in the dormitory, 17. Completeness of cleaning tools, 18. Facilities functions (bathrooms, toilets etc.),19. Water Availability (Drinking water and water for other purposes), 20. Availability of electricity and other electricity-related facilities, 21. Availability of fast Internet access (WIFI) (IT and communication facilities), 22. Infrastructure conditions and arrangements (beds, chairs, cupboards, tables, etc.), 23. Lighting in the Dormitory, 24. Security in the Dormitory environment, 25. Cleanliness / Structure the sports room, 26. Cleanliness / Arrangement of art space, 27. Laboratory equipment supplies, 28. Art and sports extracurricular services, 29. Sports equipment, 30. Hospitality Service
30 questions are grouped into 4 sections, namely: PBM services; Classroom Conditions; Dorm Condition; Arts, sports and laboratories. Results of the questionnaire distributed to cadets (44 people) obtained the evaluation results of the cadets' perception of the management of learning and infrastructure: PBM services with an average value of 27.43 (Good/Satisfactory), Classroom Conditions
with an average value of 32.36 (Good / Satisfactory), Boarding conditions with an average rating of 40.14 (Good /Satisfactory), Arts, Sports and Laboratories with an average rating of 22.46 (Good / Satisfactory).
After cadets perseption calculation process, then
calculate sgnificant test. df = 42 and alpha=0.05 so
rtable = 0.297315212 . And Reliability value = 0.960.
Reliability > 0.7, so attribute is realiable
Based on the results of calculations, all the attributes
stated are valid. and reliable so all attributes can be
used as attributes of cadets expectation. From the
calculation results the highest validity value is
Lighting Room study = 0.862 and the lowest validity
value is lecturer gives time to explain the module /
teaching material = 0.434.
The next step is to group cadets expectations into
five characteristic dimensions. Classification of
attribute cadets expectations towards learning
management services and facilities and infrastructure
standards are:
1. Dimension Tangibility : Schedule Distribution
(4), Study Room Cleanliness (9), Condition of Air
Conditioning(AC) in Study Rooms (10),
Condition of Chairs and Tables in the study room
(11), Cleanliness in the Dormitory and the
Dormitory environment (15), Condition of AC in
dormitories (16), Completeness of Cleaning
Equipment (17), Facilities Function (Bathroom,
toilet) (18), Availability of Water (Drinking
Water or water for other other purposes) (19),
Availability of Electricity and Other Electrical
Related Facilities (20), Infrastructure conditions
and arrangements (beds, chairs, cupboards, tables,
etc.) (22), Dormitory Lighting (23), Sports Room
Cleaning / Arrangement (25), Art Space Cleaning
/ Arrangement (26), Laboratory equipment
supplies (27), Sports equipment (29).
2. Dimension Realibity : Conformity of PBM
implementation (6), Learning Infrastructure
Functions such as LCD / Projector (12), Layout
Tables and Chairs and other equipment (13).
3. Dimension Responsivnes : Submitting
information on the requirements and cadets needs
(1), Administrative Services (2), Information
About Student Services (7), Arts and sports
extracurricular services (28).
4. Dimension Assurance : Provision of learning
equipment (3), Provision of student Equipment
(8), Dormitory security (24), Internet Access
(WIFI) Availability (21), Study Room Lighting
(14).
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10536
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative
Commons Attribution-NonCommercial 4.0 International License. 146
5. Dimension Empaty : Hospitality Service (30),
Lecturer gives time to explain Module / Teaching
material (5).
The importance level, the cadets satisfaction level
and the improvement ratio are obtained from the
calculation of Planning Matrix. From the table it can be obtained that the highest
average cadets satisfaction level lies in the
dimensions of reliability with a satisfaction value of
4.01 followed by the Assurance dimension with a
satisfaction value of 3.98 then on
the empaty dimension with a satisfaction value of
3.97 then on the dimension of Responsivness with a
satisfaction value of 3.95 and finally
the Tangibility dimension with a satisfaction value
of 3.92.
While from the average repair ratio, the most need to
be improved is the Tangibility dimension with the
value of Improvement Ratio = 1.28 followed by the
dimension of Responsivness, the value of
Improvement Ratio = 1.264, then the dimensions
of empaty, the value of Improvement Ratio = 1.261,
next is the dimension of Assurance, the value of
Improvement Ratio = 1.259 and the last
dimension Realibility value Improvement Ratio =
1.248.
The technical characteristics of service (X3) or the
technical response for voice of customer of Medan
AVIATION POLYTECHNIC , in this case the
Quality Assurance Unit for are : Schedule distribution
base on level and class, Sosializing hygiene culture,
Scheduling maintenance of study room, Facilities
suply comsumables, Scheduling maintenance of
boarding room, Facilities provision of campus
fasility, Maintenance of laboratory fasilities,
Facilities conduct PBM monev on a schedule,
Training to administration staff, Provide information
online, Give reward to cadets who win competition
accorging to extracurricular, Increase CCTV in the
dormitory enviroment, Increase bandwidth and
computer in campus are, Give excellent service
system training to administration staff, Make lecture
contract exposure in the first week in RPS.
After knowing the technical characteristics of service,
a technical correlation is conducted to determine the
relationship between the needs of the cadets and
management response. The relationship is classified
into 3 namely: very strong, strong and weak. The
results of the analysis show a very strong
relationship is correlation between : Schedule
Distribution and Schedule distribution base on level
and class, Condition of AC in study rooms and
Scheduling maintenance of study room, Condition of
chairs and tables in the study room and Scheduling
maintenance of study room, Condition of AC in
dormitories and Scheduling maintenance of boarding
room, Completeness of Cleaning Equipment and
Facilities suply comsumables, Facilities Function
(Bathroom, toilet and Scheduling maintenance of
study room, Availability of Water and Facilities suply
comsumables, Availability of Electricity; Other
Electrical Related Facilities and Scheduling
maintenance of study room; Scheduling maintenance
of boarding room, Laboratory equipment supplies
and Maintenance of laboratory fasilities, Conformity
of PBM implementation and Facilities conduct PBM
monev on a schedule, Learning infrastructure
functions and Scheduling maintenance of boarding
room, Submitting information on the requirements
and cadets needs and Provide information online,
Administrative Services and Training to
administration staff, Information about student
services and Provide information online, Arts and
sports extracurricular services and Give reward to
cadets who win competition accorging to
extracurricular, Provision of learning equipment and
Facilities suply comsumables, Internet Access
(WIFI) Availability and Increase bandwidth and
computer in campus area, Hospitality Service and
Give excellent service system training to
administration staff , Lecturer gives time to explain
Teaching material and Make lecture contract
exposure in the first week in RPS.
Strong relationship is correlation betweens: Study
room cleanliness and Sosializing hygiene culture,
Cleanliness in the Dormitory and the Dormitory
environment and Sosializing hygiene culture,
Condition of AC in dormitories and Facilities
provision of campus fasility, Facilities function and
Sosializing hygiene culture, Dormitory lighting and
Scheduling maintenance of boarding room, Sports
room and art space cleaning and Sosializing hygiene
culture, Laboratory equipment supplies and Facilities
provision of campus fasility, Conformity of PBM
implementation and Schedule distribution base on
level and class.
Weak relationships are correlation betweens : Sports
equipment and Facilities provision of campus
fasility.
The last step, quality of education services for
learning management standards and infrastructure
facilities can be seen in Figure 4.
From the calculation of the House of Quality Matrix
(HOQ) can be obtained the relationship between the
needs of cadets and technical descriptors according to
the highest order are: Scheduling maintenance of
study room facilities with a weight of 1.952;
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10536
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative
Commons Attribution-NonCommercial 4.0 International License. 147
Procurement of consumables with a weight of 1.177 ;
Scheduling of maintenance of dormitory facilities
with a weight of 0.664 ; Providing information on
online cadets services weighing 0.589 ; Procurement
of campus facilities checklist with a weight of 0.408
; The division of schedules by force and class with a
weight of 0.394 ; To socialize clean culture with a
weight of 0.332 ; Provide excellent service system
training to administrative staff with a weight of 0.320
; Adding lecture contract exposure in the first week
of the RPS with a weight of 0.3 ; Give rewards to
cadets who win the competition according to
extracurriculars with a weight of 0.298 ; Provide
training to administrative staff with a weight of 0.291
; Conduct PBM monev on a scheduled basis with a
weight of 0.288 ; Adding internet and computer
bandwidth in the campus area with a weight of 0.286
; Maintenance of laboratory facilities with a weight of
0.262 ; Adding CCTV in the dormitory environment
with a weight of 0.123.
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10536
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative
Commons Attribution-NonCommercial 4.0 International License. 148
Figure4. Matrix House Of Quality (HOQ)
V. CONCLUSION AND SUGGESTION
The results of the analysis and discussion, the
following conclusions are obtained:
1. There is a significant difference between the
hopes / desires of cadets and the cadets'
perceptions of learning management services and
the facilities of the Medan POLYTECHNIC
AVIATION facility.
2. Based on the importance level of service attributes
that are very important for the Medan AVIATION
POLYTECHNIC cadets there are 4 namely:
Level of importance of the Facility Function
Note relationship:
9 > very strong
3 > strong
1 > weak
LEV
EL O
F SA
TISF
AC
TIO
N
Go
al/
Targ
et
IMP
RO
VEM
ENT
RA
TIO
S
Sell
ing
Po
int
Raw
We
igh
t
No
rmal
ize
d W
eig
ht
Schedule Distribution (4) 4 0,305,00 1,28 1,5 7,67 0,03
Study Room Cleanliness (9) 4 0,095,00 1,22 1,5 7,33 0,03
Condition of Air Conditioning in Study Rooms (10) 4 0,285,00 1,19 1,5 7,14 0,03
Condition of Chairs and Tables in the study room (11) 4 0,295,00 1,26 1,5 7,59 0,03
Cleanliness in the Dormitory and the Dormitory
environment (15)
4 0,09
5,00 1,18 1,5 7,10 0,03
Condition of Air Conditioning (AC) in dormitories (16) 3 0,23 0,085,00 1,33 1,5 6,00 0,03
Completeness of Cleaning Equipment (17) 4 0,30 0,105,00 1,30 1,5 7,81 0,03
Facilities Function (Bathroom, toilet) (18) 4 0,12 0,305,00 1,26 1,5 9,43 0,04
Availability of Water (Drinking Water or water for other
other purposes) (19)
4 0,28
5,00 1,19 1,5 7,14 0,03
Availability of Electricity and Other Electrical Related
Facilities (20)
4 0,31 0,31
5,00 1,34 1,5 8,05 0,03
Infrastructure conditions and arrangements (beds,
chairs, cupboards, tables, etc.) (22)
4 0,34
5,00 1,17 1,5 8,78 0,04
Dormitory Lighting (23) 4 0,125,00 1,22 1,5 9,12 0,04
Sports Room Cleaning / Arrangement (25) 4 0,015,00 1,27 1,5 7,63 0,03
Art Space Cleaning / Arrangement (26) 4 0,015,00 1,37 1,5 8,20 0,04
Laboratory equipment supplies (27) 3 0,09 0,265,00 1,50 1,5 6,73 0,03
Sports equipment (29) 4 0,045,00 1,40 1,5 8,41 0,04
Conformity of PBM implementation (6)4
0,10 0,295,00 1,24 1,5 7,42 0,03
Learning Infrastructure Functions such as LCD /
Projector (12)
4 0,29
5,00 1,26 1,5 7,59 0,03
Layout Tables and Chairs and other equipment (13) 4 0,035,00 1,24 1,5 7,46 0,03
Submitting information on the requirements and cadets
needs (1)
4 0,29
5,00 1,26 1,5 7,59 0,03
Administrative Services (2) 4 0,290,03 5,00 1,25 1,5 7,50 0,03
Information About student Services (7)4
0,295,00 1,26 1,5 7,59 0,03
Arts and sports extracurricular services (28)
4
0,30
5,00 1,28 1,5 7,67 0,03
Provision of learning equipment (3) 4 0,32 0,11
5,00 1,36 1,5 8,15 0,04
Provision of Student Equipment (8) 4 0,28
5,00 1,20 1,5 7,21 0,03
Dormitory security (24)
4
0,12
5,00 1,26 1,5 9,48 0,04
Internet Access (WIFI) Availability (21)
4
0,29
5,00 1,23 1,5 7,37 0,03
Study Room Lighting (14) 4 0,10
5,00 1,24 1,5 7,46 0,03
Hospitality Service (30)
4
0,29
5,00 1,24 1,5 7,42 0,03
Lecturer gives time to explain Module / Teaching
material (5)
4 0,30
5,00 1,29 1,5 7,72 0,03
Total0,39 0,33 1,95 1,18 0,66 0,41 0,26 0,29 0,29 0,59 0,30 0,12 0,29 0,32 0,30
Rank 6 7 1 2 3 5 14 12 11 4 10 15 13 8 9
Giv
e e
xc
elle
nt s
erv
ice
sy
ste
m tra
inin
g to
ad
min
istra
tion
sta
ff
Ma
ke
lec
ture
co
ntra
ct e
xp
os
ure
in th
e firs
t
we
ek
in R
PS
TECHNICAL DESCRIPTOR (HOWS)
Sc
he
du
le d
istrib
utio
n b
as
e o
n le
ve
l an
d
cla
ss
So
sia
lizin
g h
yg
ien
e c
ultu
re
Sc
he
du
ling
ma
inte
na
nc
e o
f stu
dy
roo
m
Fa
cilitie
s s
up
ly c
om
su
ma
ble
s
Sc
he
du
ling
ma
inte
na
nc
e o
f bo
ard
ing
roo
m
Fa
cilitie
s p
rov
isio
n o
f ca
mp
us
fas
ility
Ma
inte
na
nc
e o
f lab
ora
tory
fas
ilities
Fa
cilitie
s c
on
du
ct P
BM
mo
ne
v o
n a
sc
he
du
le
Tra
inin
g to
ad
min
istra
tion
sta
ff
CA
DE
T R
EQ
UIR
EM
EN
T (
WH
AT
S)
Pro
vid
e in
form
atio
n o
nlin
e
Giv
e re
wa
rd to
ca
de
ts w
ho
win
co
mp
etitio
n
ac
co
rgin
g to
ex
trac
urric
ula
r
Inc
rea
se
CC
TV
in th
e d
orm
itory
en
viro
me
nt
Inc
rea
se
ba
nd
wid
th a
nd
co
mp
ute
r in
ca
mp
us
are
a
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10536
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative
Commons Attribution-NonCommercial 4.0 International License. 149
(Bathroom, toilet), Condition and infrastructure
arrangements (beds, Chairs, Cabinets, tables etc.),
Lighting in the Dormitory, Security boarding
environment.
3. From the results of the analysis of satisfaction of
cadets towards learning management services and
facilities of Medan AVIATION
POLYTECHNIC, the highest level of satisfaction
of cadets is the dimension of reliability with a
satisfaction value of 4.01, followed by the
Assurance dimension with a value of satisfaction
3.98 and then on the dimension of empaty with a
value of satisfaction of 3.97 then on the
Responsivness dimension with a satisfaction
value of 3.95 and the last is the Tangibility
dimension with a satisfaction value of 3.92.
4. From the analysis of the improvement ratio, the
increase that must be done by Medan AVIATION
POLYTECHNIC for the highest learning
management and infrastructure is the Tangibility
dimension with the value of Improvement Ratio =
1.28 followed by the dimension of Responsivity,
the value of Improvement Ratio = 1,264, then the
empathy dimension, the value of Improvement
Ratio = 1,261 hereinafter is the Assurance
dimension Improvement Ratio value = 1.259 and
the last dimension is the Reliability dimension
Improvement Ratio value = 1.248.
5. Based on the analysis using the QFD method , to
meet the cadets needs, the priority of Medan
AVIATION POLYTECHNIC service that can be
used as a Quality Standards of learning
management and facilities is the Scheduling of
maintenance of study room facilities with a
weight of 1.952 followed by the procurement of
consumables with a weight of 1.177 then
scheduling maintenance of dormitory facilities
with a weight of 0.664, providing online cadets
service information with a weight of 0.589 then
procuring a campus facility checklist with a
weight of 0.408 followed by schedule distribution
by force and class with a weight of 0.394,
socializing a clean culture with a weight of 0.332,
providing excellent service system training to
staff administration with a weight of 0.320,
adding exposure to the college contract in the first
week of the RPS with a weight of 0.3, giving
rewards to cadets who win the competition
according to extracurriculars with a weight of
0.298, providing training to administrative staff
with with a weight of 0.291, scheduled PBM
monitoring and evaluation with a weight of 0.288,
increasing internet and computer bandwidth in the
campus area with a weight of 0.286, maintenance
of laboratory facilities with a weight of 0.262 and
the last addition of CCTV in the dormitory
environment with a weight of 0.123
Suggestions for further research
1. To obtain a quality standard quality of learning
management infrastructures and facilities in
Medan AVIATION POLYTECHNIC advisable
to do a questionnaire and analysis for lecture and
staff as internal custumer.
2. To obtain quality standards beyond the SNDIKTI
Quality Standards it is recommended to use the
QFD method in setting standards for all quality
assurance pen standards with subjects that suit
your needs.
VI. REFERENCES
Abdulrahman F, I., & Handayani, A. (2017). House
Of Quality Sebagai Pengendalian Kualitas
Layanan di Lembaga Pendidikan
Muhammadiyah. Prosiding Seminar
Nasional AIMI (pp. 12-25). Jambi: AIMI. Alibeigi, M. (2015). A Robust Statistical Color Edge
Detection for Noisy Images.
SemanticScholar.
Andre A et al. (2016). Analisis Kepuasan Mahasiswa
Terhadap Fasilitas Dan Layanan Karyawan
Di Teknik Industri Universitas Diponegoro
Dengan Zone Of Tolerance Dan Quality
Function Deployment. Industrial
Engineering Online Journal, V(1).
Azizah N. I. et al., R. L. (2018, Agustus 2). Penerapan
Metode Dan Quality Function Deployment
(QFD) Dalam Memenuhi kepuasan
Konsumen Pada Industri Komponen
Otomotif. Jurnal Teknik Industri, XIX(2),
127-136. doi:10.22219
Claudia Gita, P., & Susanti, A. (2017). Analisis
Kepuasan Pelanggan Menggunakan Metode
Importance Performance Analysis Dan
Quality Function Deployment Di
Hypermarket Super Indo Tembalang .
Industrial Engineering Online Journal.,
VI(4).
Cohen, L. (2007). Quality Function Deployment:
How to Make QFD Work For You. Virginia:
R.R Donnelly Horrisonburg.
Ginting, M., & Halim, D. I. (2012, April ). Usaha
Peningkatan Kualitas Pelayanan
Perpustakaan Ukrida Dengan Metode
Serqual Dan Quality Function Deployment
(QFD). Jurnal Teknik dan Informatika
Komputer., I(2), 182-195.
SinkrOn : Jurnal dan Penelitian Teknik Informatika Volume 4, Number 2, April 2020
DOI : https://doi.org/10.33395/sinkron.v4i2.10536
e-ISSN : 2541-2019 p-ISSN : 2541-044X
This is an Creative Commons License This work is licensed under a Creative
Commons Attribution-NonCommercial 4.0 International License. 150
Hashem, H. F. (2009). Adaptive technique for human
face detection using HSV color space and
neural networks. National Radio Science
Conference. IEEE.
Hasim, & Yusuf, A. (2010). Determining Suffiency
Of Sample Size In Management urvey
Research Activities. International Journal
of Organisasional Manajemen &
Entepreneurship Development, 6(1), 119-
130.
Kusumawardhani, A., & Shafiq, N. (2018, Desember
28). Teknika : Engineering and Sains
Journal . Implementasi Metode Quality
Function Deployment (QFD) Dalam
Menganalisis Kualitas Pelayanan Pada
Bengkel Politeknik, II(2), 95-102.
Pittsburgh, U. o. (2019, February 18). University of
Library System. Retrieved from Citation
Styles: APA, MLA, Chicago, Turabian,
IEEE: APA 6th Edition:
https://pitt.libguides.com/c.php?g=12108&
p=64730
Republik Indonesia. (2018). Amandement Menteri
Riset, Teknologi, Dan Pendidikan Tinggi .
Tentang Standar Nasional Pendidikan
Tinggi, Jakarta.
Republik Indonesia. (2015). Peraturan Menteri Riset,
Teknologi, Dan Pendidikan Tinggi. Tentang
Standart Pendidikan Tinggi Nasional,
Jakarta.
Rusiadi et al., S. N. (2014). Metode Penelitian
Manajemen, Akuntansi dan Ekonomi
Pembangunan. Medan: USU Press.
Sugiono, D. (2010). Statistika Untuk Penelitian.
Bandun: Alfabeta.
Sutawijaya, A. H., & Pista, S. (2018, September 2).
Evaluasi Pelayanan Publik Produk
Hukumonline.com Untuk Mengetahuio
Kebutuhan Pelanggan Kasus PT. Justika
Publika. JDM, I(2), 32-45.
Tjiptono, F. ( 2012). Service Management
Mewujudkan Layanan Prima. Yogyakarta:
Penerbit Andi.
Tutuhatunewa, A. (2010). Aplikasi Metode Quality
Function Deployment Dalam
Pengembangan Produk Air Minum
Kemasan. ARIKA, 11-19.
Yamit, Z. (2013). Manajemen Kualitas Produk &
Jasa. (Vol. VI). Yogyakarta,
D.I.,Yogyakarta, Indonesia: Penerbit
EKONISIA.