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

[email protected]

Nining Suharyanti


West Jakarta City, Indonesia

[email protected]

Triningsih



[email protected]

Murniyati



[email protected]

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.

https://doi.org/10.33395/sinkron.v4i2.10500

mailto:[email protected]






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.




e-ISSN : 2541-2019 p-ISSN : 2541-044X


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%




e-ISSN : 2541-2019 p-ISSN : 2541-044X


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


8. Meatzza, Chesee mania with Support Value of


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:




e-ISSN : 2541-2019 p-ISSN : 2541-044X


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.




e-ISSN : 2541-2019 p-ISSN : 2541-044X


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

[email protected]

Mario Susanto


Medan, Indonesia

[email protected]

Andy Chandra


Medan, Indonesia

[email protected]




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







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)


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).


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




e-ISSN : 2541-2019 p-ISSN : 2541-044X


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.


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.




e-ISSN : 2541-2019 p-ISSN : 2541-044X


Figure 4. Overall Graph of Borax Content Classification Systemn On Tomato Sauce Using

GLCM


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.



DOI https://doi.org/10.33395/sinkron.v4i2.10481

e-ISSN : 2541-2019 p-ISSN : 2541-044X



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]




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

https://doi.org/10.33395/bsi.ac.id




e-ISSN : 2541-2019 p-ISSN : 2541-044X


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




e-ISSN : 2541-2019 p-ISSN : 2541-044X


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.


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




e-ISSN : 2541-2019 p-ISSN : 2541-044X


achieved, then the whole process will be done

sequentially.


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.


https://www.bps.go.id/



e-ISSN : 2541-2019 p-ISSN : 2541-044X


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




e-ISSN : 2541-2019 p-ISSN : 2541-044X


Figure 8 Display Symptom List

Figure 9 Display Detailed Data Symptoms

Figure 10 Disease List Display

Figure 11 Disease Detailed Data Display


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




e-ISSN : 2541-2019 p-ISSN : 2541-044X


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




e-ISSN : 2541-2019 p-ISSN : 2541-044X


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


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




e-ISSN : 2541-2019 p-ISSN : 2541-044X


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.


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.


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.




e-ISSN : 2541-2019 p-ISSN : 2541-044X



Analysis of Multi-attribute Utility Theory for

College Ranking Decision Making

Adidtya Perdana

Universitas Harapan Medan

Medan, Indonesia

[email protected]

Arief Budiman


Medan, Indonesia

[email protected]

Submitted: Oct 2, 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






e-ISSN : 2541-2019 p-ISSN : 2541-044X



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.


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




e-ISSN : 2541-2019 p-ISSN : 2541-044X



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


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


a. 80% - 100% : 4

b. 60% - 79.99% : 3

c. 40% - 59.99% : 2




e-ISSN : 2541-2019 p-ISSN : 2541-044X



d. 0% - 39.99% : 1

d. C4 = HR: Lecturer with Doctoral Degree


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




e-ISSN : 2541-2019 p-ISSN : 2541-044X



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


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




e-ISSN : 2541-2019 p-ISSN : 2541-044X



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


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




e-ISSN : 2541-2019 p-ISSN : 2541-044X



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.


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.




e-ISSN : 2541-2019 p-ISSN : 2541-044X



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.




e-ISSN : 2541-2019 p-ISSN : 2541-044X



Decision Support System For Determining the

Best College High Private Using Topsis Method

Yuyun Dwi Lestari


Medan, Indonesia

[email protected]

Mardiana


Medan, Indonesia

[email protected]

Submitted: Oct 3, 2019



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).






e-ISSN : 2541-2019 p-ISSN : 2541-044X



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)


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).


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:




e-ISSN : 2541-2019 p-ISSN : 2541-044X



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



a. 0% : 4

b. 0.1% - 0.99% : 3

c. 1% - 8% : 2

d. > 8% : 1



a. 80% - 100% : 4

b. 60% - 79.99% : 3

c. 40% - 59.99% : 2

d. 0% - 39.99% : 1



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


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


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




e-ISSN : 2541-2019 p-ISSN : 2541-044X



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


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




e-ISSN : 2541-2019 p-ISSN : 2541-044X



9 C9 0,75

10 C10 0,25

Next, calculate the weight of each data used:

TABLE IV. NORMALIZATION MATRIX RESULTS

No Data


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




e-ISSN : 2541-2019 p-ISSN : 2541-044X



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


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




e-ISSN : 2541-2019 p-ISSN : 2541-044X



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.





Toddler Nutrition Status. SinkrOn, 4(1), 19.


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.





e-ISSN : 2541-2019 p-ISSN : 2541-044X



Pears Classification Using Principal

Component Analysis and K-Nearest Neighbor

Moh. Arie Hasan

Sekolah Tinggi Mamajemen Informatika dan Komputer

Nusa Mandiri

Jakarta, Indonesia

[email protected]

Arief Setya Budi

Sekolah Tinggi Mamajemen Informatika dan

Komputer Nusa Mandiri

Jakarta, Indonesia

[email protected]




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






e-ISSN : 2541-2019 p-ISSN : 2541-044X



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.


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




e-ISSN : 2541-2019 p-ISSN : 2541-044X



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).


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)




e-ISSN : 2541-2019 p-ISSN : 2541-044X



(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.


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




e-ISSN : 2541-2019 p-ISSN : 2541-044X



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:




e-ISSN : 2541-2019 p-ISSN : 2541-044X



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


Classification of Red Pears


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.




e-ISSN : 2541-2019 p-ISSN : 2541-044X



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


22


23


24


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




e-ISSN : 2541-2019 p-ISSN : 2541-044X



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)


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.


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.




e-ISSN : 2541-2019 p-ISSN : 2541-044X



Analysis K-Nearest Neighbor Algorithm for

Improving Prediction Student Graduation Time

Rizki Muliono

Universitas Medan Area

Medan, Indonesia

[email protected]

Juanda Hakim Lubis


Medan, Indonesia

[email protected]

Nurul Khairina


Medan, Indonesia

[email protected]

Submitted: Jan 30, 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.







e-ISSN : 2541-2019 p-ISSN : 2541-044X



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)




e-ISSN : 2541-2019 p-ISSN : 2541-044X



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










12 158110020 Incorrect Correct 0









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


Classification


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




e-ISSN : 2541-2019 p-ISSN : 2541-044X



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.


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




e-ISSN : 2541-2019 p-ISSN : 2541-044X



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




e-ISSN : 2541-2019 p-ISSN : 2541-044X



Apriori Algorithm on Car Rental Analysis with

The Most Popular Brands

Leo Fernando Panjaitan

STMIK Nusa Mandiri

Jakarta, Indonesia

[email protected]

Yopi Handrianto

Universitas Bina Sarana Informatika

Jakarta, Indonesia

[email protected]

Achmad Nurhadi


Jakarta, Indonesia

[email protected]




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







e-ISSN : 2541-2019 p-ISSN : 2541-044X



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..


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.




e-ISSN : 2541-2019 p-ISSN : 2541-044X



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




e-ISSN : 2541-2019 p-ISSN : 2541-044X



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)



And for the support value of 3 items obtained from the following formula:

𝑺𝒖𝒑𝒑𝒐𝒓𝒕(𝐀, 𝐁, 𝐂)∑ 𝐓𝐫𝐚𝐧𝐬𝐚𝐜𝐭𝐢𝐨𝐧 𝐂𝐨𝐧𝐭𝐚𝐢𝐧𝐢𝐧𝐠 𝐀, 𝐁 𝐝𝐚𝐧 𝐂


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


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:




e-ISSN : 2541-2019 p-ISSN : 2541-044X



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


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


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%




e-ISSN : 2541-2019 p-ISSN : 2541-044X



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%


B. 2 itemset combination

Formation of Combinations The support values of 2 items are obtained by the following formula:



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%




e-ISSN : 2541-2019 p-ISSN : 2541-044X



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%


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:




e-ISSN : 2541-2019 p-ISSN : 2541-044X



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%


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


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%


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



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%


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%


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.




e-ISSN : 2541-2019 p-ISSN : 2541-044X




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.


Irfiani, E. (2019). Application of Apriori Algorithms

to Determine Associations in Outdoor Sports

Equipment Stores. SinkrOn, 3(2), 218.


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.




e-ISSN : 2541-2019 p-ISSN : 2541-044X



Pythagoras Tree Applied For Determined

Instagram Usage Habit Decision

Erlin Windia Ambarsari

Universitas Indraprasta PGRI

Jakarta, Indonesia

[email protected]

Herlinda

Universitas Indraprasta PGRI

Jakarta, Indonesia

[email protected]

Submitted: Mar 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).






e-ISSN : 2541-2019 p-ISSN : 2541-044X



(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.


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)


Classification with regression method for constructing Pythagoras Tree to Instagram Habit Usage datasets have several steps as follows:




e-ISSN : 2541-2019 p-ISSN : 2541-044X



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.


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




e-ISSN : 2541-2019 p-ISSN : 2541-044X



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 <




e-ISSN : 2541-2019 p-ISSN : 2541-044X



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.


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




e-ISSN : 2541-2019 p-ISSN : 2541-044X



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.


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




e-ISSN : 2541-2019 p-ISSN : 2541-044X



Selection of Outstanding Lecturers with Simple

Additive Weighting Method

Embun Fajar Wati


West Jakarta, Indonesian

[email protected]

Istikharoh

STIKOM Cipta Karya Informatika

West Jakarta, Indonesian

[email protected]

Tuslaela

STMIK Nusa Mandiri

Central Jakarta, Indonesian

[email protected]




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







e-ISSN : 2541-2019 p-ISSN : 2541-044X



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).


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




e-ISSN : 2541-2019 p-ISSN : 2541-044X



(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.


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.


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




e-ISSN : 2541-2019 p-ISSN : 2541-044X



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




e-ISSN : 2541-2019 p-ISSN : 2541-044X



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.


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,




e-ISSN : 2541-2019 p-ISSN : 2541-044X



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.




e-ISSN : 2541-2019 p-ISSN : 2541-044X



Expert System Detects Problems of Inkmaker

Machine And Mixer With Forward Chaining

M. Sinta Nurhayati

Bina Sarana Informatika University Jakarta

Jakarta, Indonesia

[email protected]

Rachmat Hidayat

Bina Sarana Informatika University Jakarta

Jakarta, Indonesia

[email protected]




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






e-ISSN : 2541-2019 p-ISSN : 2541-044X



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.


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




e-ISSN : 2541-2019 p-ISSN : 2541-044X



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.


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.




e-ISSN : 2541-2019 p-ISSN : 2541-044X



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




e-ISSN : 2541-2019 p-ISSN : 2541-044X



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


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




e-ISSN : 2541-2019 p-ISSN : 2541-044X



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




e-ISSN : 2541-2019 p-ISSN : 2541-044X



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


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.


ticle/view/113/75

Priyadi, Y. (2015). Kolaborasi SQL & ERD dalam

Implementasi Database. Liris.

Pudji W, E. I. (2012). Preventive Maintenance PT




e-ISSN : 2541-2019 p-ISSN : 2541-044X



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




e-ISSN : 2541-2019 p-ISSN : 2541-044X



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

[email protected]

Henny Destiana

UBSI Jakarta

Jakarta, Jl. Kamal Raya No.18, RT.6/RW.3,

Cengkareng, Kota Jakarta Barat

[email protected]




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






e-ISSN : 2541-2019 p-ISSN : 2541-044X



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).


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)




e-ISSN : 2541-2019 p-ISSN : 2541-044X




Figure 1 research stages


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%




e-ISSN : 2541-2019 p-ISSN : 2541-044X



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%


No Menu Table


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


1. Float 22/30 73%

2. Ice Lemon Tea 13/30 43%

3. Cappucinno 13/30 43%


5. Hot/Ice Coffe Latte 5/30 16%


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%





e-ISSN : 2541-2019 p-ISSN : 2541-044X



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:




e-ISSN : 2541-2019 p-ISSN : 2541-044X



TABLE 4. Value of Support 2 Itemset

No Menu Table


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


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%


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


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%




e-ISSN : 2541-2019 p-ISSN : 2541-044X



No Menu Table


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%


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%




e-ISSN : 2541-2019 p-ISSN : 2541-044X



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%


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.


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).




e-ISSN : 2541-2019 p-ISSN : 2541-044X



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.


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




e-ISSN : 2541-2019 p-ISSN : 2541-044X



Optimization of Sugar Salt Fat in the Human

Body Using Genetic Algorithm

Preddy Marpaung1

STMIK Pelita Nusantara

Medan, Indonesia

[email protected]

Arjon Samuel Sitio2


Medan, Indonesia

[email protected]

Anita Sindar3


Medan, Indonesia

[email protected]




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







e-ISSN : 2541-2019 p-ISSN : 2541-044X


(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.


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




e-ISSN : 2541-2019 p-ISSN : 2541-044X


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.


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




e-ISSN : 2541-2019 p-ISSN : 2541-044X



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


S4 Vegetables And


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


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




e-ISSN : 2541-2019 p-ISSN : 2541-044X


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















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





Parent 4 25737.27







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



Parent 11 63759.45

Parent 13 64829.42

Parent 2 67460.21

Parent 5 73638.29


Parent 14 74925.02


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




e-ISSN : 2541-2019 p-ISSN : 2541-044X


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.




e-ISSN : 2541-2019 p-ISSN : 2541-044X



Decision Support System for Achieving

Scholarship Selection by Using Profile

Matching Method

Rani Irma Handayani1st


[email protected]

Triningsih2nd


[email protected]

Melia Putri 3rd


[email protected]


Accepted: 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







e-ISSN : 2541-2019 p-ISSN : 2541-044X



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)


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




e-ISSN : 2541-2019 p-ISSN : 2541-044X



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



6 3 2,5 Individual competencies are

over 3 levels / level




e-ISSN : 2541-2019 p-ISSN : 2541-044X





8 4 1,5 Individual competence is

over 4 levels / level



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%




e-ISSN : 2541-2019 p-ISSN : 2541-044X



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




e-ISSN : 2541-2019 p-ISSN : 2541-044X



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.




e-ISSN : 2541-2019 p-ISSN : 2541-044X



Dog Disease Expert System Using

Certainty Factor Method

Linda Marlinda

STMIK Nusa Mandiri

Jakarta, Indonesia

[email protected]

Widiyawati

STMIK Bani Saleh

Bekasi, Indonesia

[email protected]

Wahyu Indrarti


Jakarta, Indonesia

[email protected]

Reni Widiastuti


Jakarta, Indonesia

[email protected]




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








e-ISSN : 2541-2019 p-ISSN : 2541-044X



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).


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




e-ISSN : 2541-2019 p-ISSN : 2541-044X



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.


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


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




e-ISSN : 2541-2019 p-ISSN : 2541-044X



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


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


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




e-ISSN : 2541-2019 p-ISSN : 2541-044X



"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


= 0.9475 + 0.72 * (1 - 0.9475)

= 0.9853


= 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


G23 AND G24

P7


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




e-ISSN : 2541-2019 p-ISSN : 2541-044X



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


lethargy occurred 0,4


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


Fever 0,2

complaints 0,8

Shivering 0,6

dirofilaria

immitis (P8)

Cough 0,2

complaints 0,4

out of breath 0,4


Unstable body

temperature 0,9

Display Interface

Display questionnaires for symptoms of canine disease




e-ISSN : 2541-2019 p-ISSN : 2541-044X




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.


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




e-ISSN : 2541-2019 p-ISSN : 2541-044X



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




e-ISSN : 2541-2019 p-ISSN : 2541-044X



Support Vector Machine Parameter

Optimization to Improve Liver Disease

Estimation with Genetic Algorithm

Hani Harafani

STMIK Nusa Mandiri

Jakarta

[email protected]




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





e-ISSN : 2541-2019 p-ISSN : 2541-044X



(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.


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




e-ISSN : 2541-2019 p-ISSN : 2541-044X



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.


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


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




e-ISSN : 2541-2019 p-ISSN : 2541-044X



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.




e-ISSN : 2541-2019 p-ISSN : 2541-044X



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


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




e-ISSN : 2541-2019 p-ISSN : 2541-044X



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




e-ISSN : 2541-2019 p-ISSN : 2541-044X



P(T<=t) two-tail 0,034121


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.


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




e-ISSN : 2541-2019 p-ISSN : 2541-044X



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.


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.


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.


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




e-ISSN : 2541-2019 p-ISSN : 2541-044X



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]




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







e-ISSN : 2541-2019 p-ISSN : 2541-044X



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.


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).


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)




e-ISSN : 2541-2019 p-ISSN : 2541-044X



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.


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.




e-ISSN : 2541-2019 p-ISSN : 2541-044X



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




e-ISSN : 2541-2019 p-ISSN : 2541-044X



(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




e-ISSN : 2541-2019 p-ISSN : 2541-044X



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



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



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



TABLE XV. PREFERENCE VALUE

CALCULATION RESULTS




e-ISSN : 2541-2019 p-ISSN : 2541-044X



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.


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.




Toddler Nutrition Status. SinkrOn, 4(1), 19.


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


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.


Surya, C., & Wahyu, A. (2020). Sistem Informasi




e-ISSN : 2541-2019 p-ISSN : 2541-044X



Perhitungan Poin Pelanggaran Siswa

Menggunakan Metode Simple Additive

Weighting (SAW) (Studi Kasus Di SMK As-

Shofa Kabupaten Tasikmalaya). Jurnal

TEKNOINFO, 14(1), 59.





e-ISSN : 2541-2019 p-ISSN : 2541-044X



Analysis of Taklinear Performance and Integer

Linear Programming Models in Nurses

Scheduling Problems

Junerdi Nababan

University Sumatera Utara

Medan, Indonesia

[email protected]

Tulus


Medan, Indonesia

[email protected]

Zakarias Situmorang


Medan, Indonesia

[email protected]




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







e-ISSN : 2541-2019 p-ISSN : 2541-044X



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,




e-ISSN : 2541-2019 p-ISSN : 2541-044X



(∑ ∑ 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


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




e-ISSN : 2541-2019 p-ISSN : 2541-044X



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




e-ISSN : 2541-2019 p-ISSN : 2541-044X



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




e-ISSN : 2541-2019 p-ISSN : 2541-044X



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.


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




e-ISSN : 2541-2019 p-ISSN : 2541-044X



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.




e-ISSN : 2541-2019 p-ISSN : 2541-044X



Evaluation of the Satisfaction of Users of

Weather Forecast Systems with the Service

Quality Method

Eva Rianti

Universitas Putra Indonesia YPTK Padang

Padang, Indonesia

[email protected]

Syafrika Deni Rizky


Padang, Indonesia

[email protected]

Fariz Haris Nugraha


Padang, Indonesia

[email protected]




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.







e-ISSN : 2541-2019 p-ISSN : 2541-044X




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




e-ISSN : 2541-2019 p-ISSN : 2541-044X



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.


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.


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




e-ISSN : 2541-2019 p-ISSN : 2541-044X



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




e-ISSN : 2541-2019 p-ISSN : 2541-044X



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




e-ISSN : 2541-2019 p-ISSN : 2541-044X



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




e-ISSN : 2541-2019 p-ISSN : 2541-044X



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




e-ISSN : 2541-2019 p-ISSN : 2541-044X



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


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


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.




e-ISSN : 2541-2019 p-ISSN : 2541-044X



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:




e-ISSN : 2541-2019 p-ISSN : 2541-044X



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




e-ISSN : 2541-2019 p-ISSN : 2541-044X



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.




e-ISSN : 2541-2019 p-ISSN : 2541-044X



Analysis of Cadets Satisfaction to Medan

Aviation Polytechnic Service Using Quality

Function Deployment (QFD) Method Liber Tommy Hutabarat

POLTEKBANG MEDAN

Medan, Indonesia

[email protected]

Lisda Juliana Pangaribuan

AMIK MBP

Medan, Indonesia

[email protected]




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






e-ISSN : 2541-2019 p-ISSN : 2541-044X



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.


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




e-ISSN : 2541-2019 p-ISSN : 2541-044X



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




e-ISSN : 2541-2019 p-ISSN : 2541-044X



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)


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.




e-ISSN : 2541-2019 p-ISSN : 2541-044X



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.


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).




e-ISSN : 2541-2019 p-ISSN : 2541-044X



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;




e-ISSN : 2541-2019 p-ISSN : 2541-044X



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.




e-ISSN : 2541-2019 p-ISSN : 2541-044X



Figure4. Matrix House Of Quality (HOQ)


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




e-ISSN : 2541-2019 p-ISSN : 2541-044X



(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.




e-ISSN : 2541-2019 p-ISSN : 2541-044X



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.


INFORMATICS ENGINEERING JOURNALS & RESEARCH

Documents

Transcript of INFORMATICS ENGINEERING JOURNALS & RESEARCH