Project Research Proposal

download Project Research Proposal

of 23

Transcript of Project Research Proposal

  • 8/6/2019 Project Research Proposal

    1/23

    KWAZULU UNIVERSITY

    SCHOOL OF COMPUTING AND INFORMATICS

    BACHELOR SCIENCE IN INFORMATION

    TECHNOLOGY

    PROJECTRESEARCH PROPOSAL.

    TITLE:

    PERSONALIZED ADAPTIVE LEARNING MODEL

    DEVELOPED AS A SERVICE AND DEPLOYED UNDER

    INTERMITTENT INTERNET CONNECTION CONDITIONS.

    NAME: JANE PHILLIP

    REGNO: CO1/4038/2000

    SUPERVISOR: MR. PETER NDEGWA

  • 8/6/2019 Project Research Proposal

    2/23

    Key terms

    API Application Programming Interface

    DLL-Dynamic Link Library

    KNN-K-nearest neighbor

    LMS-Learning Management System

    HTTP-Hypertext Transfer Protocol

    Master model and master API are used interchangeably.

    Learner model, student model and client model are used interchangeably.

  • 8/6/2019 Project Research Proposal

    3/23

    CHAPTERONE

    INTRODUCTION AND BACKGROUND

    At present, E-learning systems are widely introduced in education to enhance learning by making

    available learning materials for learners. There are various systems developed and implemented

    for learners. However, most of these systems are static and inflexible. These traditional systems

    are developed on the basis of One-size-fits -all (Blochl et al, 2003). They have been designed

    and developed to cater for and /or to serve all categories of learners. They do not take care of

    individual learning abilities and performances. They do not therefore present them (learners)

    with only relevant information but instead all information (learning materials) from the basics of

    the courses to the advanced concepts are availed to learners. This is done regardless of the state

    of knowledge the learner has acquired. This wholesome information given to learners without

    considering their needs makes them to have difficulties in perceiving the information and using

    it during answering the quizzes. These systems are also developed to work under constant

    internet connections (web applications that must be online to function). The learners can

    therefore learn only when online. These concepts are vivid in all E-learning systems which are

    deployed by learning institutions and even organizations that train their staff and clients using

    Learning Management Systems (LMS). This has been extended to firms that provide online

    recruitments. In all these situations, the systems must be online (either on internet or intranet).

    In the current modern world, although internet accessibility has increased significantly, still

    most of the areas do not have access to it. This is especially witnessed in rural areas where most

    Internet Service Providers (ISPs) have not invested in infrastructure to facilitate the internet

    connectivity of the areas in question. This limitation locks out many people who would

    otherwise have wanted to participate in this learning process through the internet. This category

    of people can therefore learn only by being present at the learning centre. Moreover, even in

    areas where internet is available, it is not always a hundred percent (100%) uptime. During theseinternet downtime times, learning does not take place unless there is a physical instructor and

    learner at the same venue or the learner has to wait for internet connection to be established. The

    time for connection reestablishment varies and other factors might even prolong it. Furthermore,

  • 8/6/2019 Project Research Proposal

    4/23

    the LMSs that have been developed and distributed for use are installed as applications and the

    whole package must be in place for implementation and subsequent usage by the audiences.

    This research proposes to overcome the above challenges by developing a personalized adaptive

    learning model especially for learning institutions (though it can be used by any learning group).

    This model will be deployed as a service as opposed to many available Learning Management

    Systems (LMS) applications. The service will be packaged as a Dynamic-Link Library (DLL).

    A dynamic-link library (DLL) is an executable file that acts as a shared library of functions.

    Dynamic linking provides a way for a process to call a function that is not part of its executable

    code. The executable code for the function is located in a DLL, which contains one or more

    functions that are compiled, linked, and stored separately from the processes that use them.

    DLLs also facilitate the sharing of data and resources. Multiple applications can simultaneously

    access the contents of a single copy of a DLL in memory.

    Dynamic linking differs from static linking in that it allows an executable module (either a .dll or

    .exe file) to include only the information needed at run time to locate the executable code for a

    DLL function. In static linking, the linker gets all of the referenced functions from the static link

    library and places it with your code into your executable file.

    Using dynamic linking instead of static linking offers several advantages. DLLs save memory,

    reduce swapping, save disk space, upgrade easier, provide after-market support, provide a

    mechanism to extend the MFC (Microsoft Foundation Classes) library classes, support multi-

    language programs, and ease the creation of international versions

    (http://msdn.microsoft.com/en-us/library/1ez7dh12.aspx ).

    The service will integrate with any Learning management system application and will have the

    advantage of being deployed both online and offline (under intermittent internet connection

    conditions). We intend to design a framework in the form of Application Programming Interface

    (API) that can be integrated to any learning management system and used to classify learners in

    various categories as defined by the model.

  • 8/6/2019 Project Research Proposal

    5/23

    An application programming interface (API) is an interface implemented by a software program

    that enables it to interact with other software. It facilitates interaction between different software

    programs similar to the way the user interface facilitates interaction between humans and

    computers. An API is implemented by applications, libraries, and operating systems to determine

    their vocabularies and calling conventions, and is used to access their services. It may include

    specifications for routines, data structures, object classes, and protocols used to communicate

    between the consumer and the implementer of the API.

    (http://en.wikipedia.org/wiki/Application_programming_interface ).

    In this research we intend to use K-Nearest neighbor algorithm to classify new learners. K-

    nearest neighbor algorithm (KNN) is part of supervised learning that has been used in many

    applications in the field of data mining, statistical pattern recognition and many others.

    KNN is a method for classifying objects based on closest training examples in the feature space.

    Our feature space is the range of scores that make up a class.

    An object is classified by a majority vote of its neighbors. In our case, the majority vote is the

    average score per class. A new learners score will be compared with all average scores of all

    classes and classified in the class whose average score is closest to the new learner. K is always a

    positive integer. The neighbours are taken from a set of objects for which the correctclassification is known. In our research, the neighbours are members of a class whose scores are

    close to each other. (http://www.codeproject.com/KB/recipes/K_nearest_neighbor.aspx)

    Problemstatement

    Online status

    Server installed with LMS

    LMS provides

    learning interface

    and has database

    with learning

    material

  • 8/6/2019 Project Research Proposal

    6/23

    The available learning models assume that every learner is always online or has constant access

    to internet. This fact is not always true as most areas especially the rural areas and even urban

    areas have not been covered with internet access infrastructure. LMSs therefore functions under

    this assumption locking out most prospective learners. In the case that internet connection is

    established, the learners are bombarded with all types of information without considering their

    level of knowledge.

    This has been shown in the diagram above where learning will only take place when there is

    internet connection and even so, there is no classification of learners. In the event that there is no

    internet access, no learning takes place.

    Existing LMSs assume same level of knowledge of all students thus necessitating them to learn

    even what they already have knowledge in. It is important to classify students according to their

    level of knowledge and avail learning material basing on this level of knowledge.

    Objectivesofthisresearch:

    y To develop an adaptive learning model to support learning under conditions ofintermittent internet connections. This model will be an API service and emphasis will be

    laid on how the API will be implemented so that it can be integrated to an LMS to

    enhance learning and classification.

    y To classify students according to the knowledge acquired and performance in quizzes.We shall then use K nearest neighbor (KNN) classification algorithm to correctly classify

    new students after having studied their initial knowledge and performance in quizzes.

    y To update the students profiles depending on their acquired knowledge, performances inquizzes, classification and thereby availing relevant learning materials. This will be testedboth online and offline.

  • 8/6/2019 Project Research Proposal

    7/23

    Research questions andexpectedoutcomes:

    y Does the model support learning under intermittent internet connections (online andoffline)? The success of this will be determined by finding out if learning is able to take

    place under both conditions and students performance rated as determined by their scores

    after taking the quizzes.

    y Does the model classify learners and is the KNN algorithm helping in learnersclassification? This will be rated as successful if learners will be classified in the various

    categories defined and if KNN algorithm concept will be used to classify new learners as

    per new learners performance against already classified ones.

    y Is the model able to work as an API service and is it able to be integrated to any LMS?This will be determined if the model will be independent of any LMS and would be

    distributed as a service to be installed and then used with other available LMS or simple

    developed learning system.

    y Are the learners profiles getting updated with relevant learning material? The success ofthis fact will be determined if each category of learners get learning materials relevant to

    their level of knowledge and class.

    Research context andsetup:

    This research is about using adaptive individualized learning model to be implemented in both

    learning institutions and any other organization that want to train their staff and clients and

    would like to classify the learners so that they (learners) can be provided with relevant learning

    information. Once the learners have been classified into categories, learners in one category will

    be provided with similar learning information. K-nearest neighbor algorithm will be used to trainthe model so that any new learner will be classified to any of the identified classes. The

    classification will be done based on the scores of the initial test done by the learner and the

  • 8/6/2019 Project Research Proposal

    8/23

    personal information the learner provides in regards to any prior knowledge he or she might have

    acquired in reference to the course of study. The model will have a questionnaire that will be

    filled by the new learner. The result of this questionnaire will be used to determine any prior

    knowledge the new learner has and will be used to classify him or her to the initial class. The

    functionality of the model will be tested when learning is both online and offline. The learners

    will be presented with a course in Database Management System. The course will have various

    levels of information; starting from preliminaries, introductory, middle level and advanced

    course concepts. These concepts will be presented to learners depending on their class or level of

    knowledge as determined by their assessment scores. Learners should be able to use the model to

    study the course and be able to attempt the questions given at the end of each section. Quizzes

    will be availed to the learners at the end of each section and scores recorded which will be used

    in subsequent classification.

    The learners will have an option of not reading the course content of each section by choosing to

    answer questions only. This fact will also be taken into consideration when classifying the

    learner depending on the scores. The quizzes will be of multiple choice answers to be correctly

    chosen. In the event that the learner decides not choose to go directly to the questions, it will be

    assumed that he or she has chosen to read the section course materials and the time taken to do

    so will be calculated and also used during classification. The time taken by the learner to read the

    section course material will be determined as follows: A mandatory button to be clicked by the

    learner at the start of the reading will be provided. To answer the questions, another mandatory

    button will be provided. The time taken between clicking the two buttons will be assumed to be

    the time taken to read the course content. If the learner clicks the second button only, it will be

    assumed that he or she has not read the course materials and treated appropriately.

    At the end of the learning, students should be rated and classified depending on the level of

    knowledge acquired and mastered as exhibited by scores in the quizzes.

    The learning materials will reside in a central remote database (in the master API) which will be

    downloaded to the learner model (client API) when online. Each learner will have a learner

    model which will be connecting to the remote master model while online for profile updates

  • 8/6/2019 Project Research Proposal

    9/23

    (classification and download of relevant learning materials) and can be used offline. The

    learners profiles will update themselves by downloading any new available materials from the

    master API and present to the learner the relevant learning material as determined by the

    learners level or class. The remote master model which will be accessed concurrently by many

    learners will be updated with all the changes in the learner models either from offline models at

    connection time and online ones.

    The key concept of this model to have an independent API that will be implemented as a service

    using DLLs functions. This will be packaged together for use by any interested learning

    institution or organization. A user interface which can be any LMS should be able to be

    integrated to the API service and be used as a proof of the concept.

    For the purposes of this research, LMS and API will share the same database and hence learning

    materials. Normally LMSs have their own databases containing their learning information and

    interfaces for user interaction with the information. If this API is to function independently, it

    will have its own database and interface.

    Audienceofthisresearch and how themodel is applicable to them.

    The key target groups are learning institutions and organizations who would like to train their

    staff and clients. The success of the model should be employed by these institutions to help them

    do supervised learning where new learners are classified based on the already classified ones.

    The learning materials is uploaded on the master model and the new version of the learning

    materials is updated to client models whenever they get online and connection is established to

    the master model. The model being a service in the form of an API can be distributed to the

    audiences who will in turn install it and integrate with their preferred LMS.

    The model after installation and integration will help save on time for physical training by the

    institutions and the learners can also learn at their own pace. In addition, the relevant learningmaterials according to the class will be availed automatically by the model.

  • 8/6/2019 Project Research Proposal

    10/23

    The relationship between time taken to study the course and the scores from the quizzes will be

    determined and this help the institutions to decide whether there will be need to set some time

    limits for certain course materials or not.

    The other benefit will be that the institutions will have an option of integrating the API to an

    LMS or not. This will immensely reduce the cost of implementation and maintenance.

  • 8/6/2019 Project Research Proposal

    11/23

    CHAPTERTWO

    LITERATUREREVIEWAND THEORY

    E-learning has been recognized as the new wave in education. It allows learners to study with

    unlimited time and space as 24 x 7 which is cost effective. Further, it is considered as a key

    strategy for increasing human capital in the knowledge society. It can be used in the organization

    for in-house training. Despite the advantages of e-learning, this system has not been well

    designed to classify learners and relevant learning materials customized for the classes or

    categories. There are many comments about the one-size-fits-all philosophy, which results in

    too much information for users and lack of personalization (Blochl et al, 2003).

    An adaptive e-learning system has been proposed by Blochl et.al (2003). This system applied the

    user-centric approach to improve its usability and acceptance by users. E-learner requirements

    are introduced to the system including user skills, learning styles and learning strategy and user

    profile (Blochl et at, 2003). In this system the user learning activities are observed and combined

    into a user profile. Also, the e-learning system is adjusted according to the dynamic user profile.

    Component technologies and artificial intelligence were used to deliver e-learning. These

    components include: Pedagogy agents, Interactivity level, Quality of feedback, Control

    strategies, Tutorial remediation and Student model.

    Pedagogy agents are used for integrating the behavior of users and e-learning components of the

    system. Pedagogy agents can be used to check student participation, track student progress

    through task procedures and address students errors. Intelligent agents can be used as tools for

    feedback. User performance during instruction and tutoring should be analyzed to monitor

    learning. Control strategies, planning for content and delivery strategies should be based on

    learner knowledge, concept structure such as curriculum. Tutorial remediation is the component

    responsible for selecting appropriate actions to be performed to accomplish a pedagogy task.

    Student model can be used to render individualized instruction in the system. Students

  • 8/6/2019 Project Research Proposal

    12/23

    instructional activities can be filtered, analyzed and sorted based on individual profile. This

    system is proposed to be adapted to changing knowledge requirements of user, be interactive and

    provide regular access to resource materials (Atolagbe, 2002).

    Web mining techniques have been used to build recommended agent for e-learning systems. This

    agent recommends activities to a learner based on his/her access history. The recommendation

    should be an on-line activity including doing an exercise, providing messages on conferencing

    systems, running an on-line simulation, or web resources. This agent is claimed to improve

    course material navigation and assist the on-line learning process (Zaiane, 2002). By observing

    user typing events, behaviors on studying lessons on web browser, tasks and examples, error

    made by users, and debugging events on the editor, the agent understands user behaviours

    (Mungunsukh, and Cheng, 2002).

    Chen et. al (2004) proposed a personalized e-learning using the item Response Theory. This

    system combines course material difficulty and learner ability as input to provide individual

    learning paths for learners. The item characteristic function, the maximum likelihood estimation

    (MLE), is applied to estimate learner ability based on explicit learner feedback.

    From what other scholars have done as shown in this review, it is evident that all the learning

    systems and / or models as developed as applications which are deployed and used by end users.

    The end users must always be online to do the learning. There has been no attempt to use any

    machine learning algorithms to classify the learners.

    Web mining techniques proposed by Mungunsukh, and Cheng (2002), uses data already stored in

    a database to do projections. We are proposing a dynamic way of classifying, where new

    members are classified as they come and use the model. Adaptive learning as proposed by Blochl

    and his colleagues just remains a proposal and they never showed ways of implementing their

    proposal. We are however, proposing a tangible method we shall use in implementing our

    proposal.

    In this research, we propose to develop a model which will be an API packaged as a DLL and

    will be used as a service by other LMS developers and users. It will be used either independently

  • 8/6/2019 Project Research Proposal

    13/23

    or integrated with other LMS. We shall have a master API which will be installed in a server to

    be accessed by clients APIs which will be remotely located. The client APIs will have two

    versions. The DLLs for offline learning and HTTP protocol for online learning.

  • 8/6/2019 Project Research Proposal

    14/23

    CHAPTERTHREE

    RESEARCHMETHODOLOGY

    y We shall design a questionnaire which we will use to capture the new learners initialgeneral knowledge of the course. The questionnaire will also find out if the learner has

    done the mandatory subjects that relates to the unit to be learned. This questionnaire will

    be part of the client model and will be filled by the learner on the initial access of the

    model. The questions will be of multiple choices. The model will score the learner based

    on the information he or she provides (as answered in the questionnaire). The learner

    model will also have a function for classifying the learners into the appropriate class.

    y We shall design the course content and calibrate it into different sections starting withintroductory notes, followed by middle level notes and more advanced notes. The course

    to be leaned, tested and whose results will be used to classify the learners will be

    Database Management System.

    y We shall design the MySql database with indexes and upload the course content. Theinformation to be learned will be indexed. We shall index the information to enhance

    easy access and management. Indexing is an efficient way to manage information stored

    in a database. The model will have pointers to these indexes so that after a learner hasbeen classified, appropriate pointers are availed to him or her. The appropriate pointers

    will point to the indexes with relevant information for the next section of learning as

    deemed appropriate. The database will reside in both master and learner models. The

    master models database will have all the information to be learned. In addition it will

    have updated profiles for all learners. The learners model database will have a

    replication of the information to be learned which is residing in the master models

    database. In addition, it will also have the information about the individuals profile. This

    individual profiles information will be updated to the master models database whenever

    there will be internet connection.

  • 8/6/2019 Project Research Proposal

    15/23

    y We shall design and develop the master DLL API with the database embedded into it. Weshall develop the model using Java programming language and package it as Java

    archives (JAR). This will then be used as a service. The user interface to interact with this

    DLL will be presented by the LMS in the event that there is an LMS. We shall do this so

    that we alienate the API from the user interface. This will mean that the API will be used

    with any user interface including known LMSs to function. We shall be able to distribute

    to any interested user to integrate into their preferred LMS or user interface. The DLL

    will have functions to collect information about the learner, inbuilt timer to determine the

    time taken by the learner to read the course section materials and answer the section

    quizzes. The scores and the time will be given to another function which will classify the

    learner and give the learner the relevant learning material.

    y We shall design and develop a client API with function to connect to the master APIwhenever there is internet connection and download the latest materials from the master

    models database into the client models database. This API with have HTTP for online

    connection and will also be packaged as a DLL to be used while offline. There will be an

    inbuilt function which will be sensing internet connection, determining connection to the

    master model (server) and hence downloading the updates. The model will also have a

    classifier function to classify learner and update his or her profile while offline. This

    updated profile will be uploaded to the master model.

    y We shall develop a simple LMS to test and prove the concept using a few learners. Weshall develop a simple user interface representative of an LMS, integrate it to the API and

    test the functions of the API. We shall then determine if the learning and classification is

    able to take place.

    y We shall do classification depending on the knowledge acquired as shown in quizzesscores and by use of k nearest neighbor algorithm. The inbuilt function will be able to

    classify learners. The classes will be predetermined and each class will have a maximum

    and minimum set of possible range of scores. A class mean, which is the average of the

  • 8/6/2019 Project Research Proposal

    16/23

    minimum and maximum set score per class, shall be calculated. This class mean shall

    however be recalculated to include members subsequently classified. The absolute

    differences between the scores of individual learners and classes means will be

    calculated and the inverse of the results found. K-nearest neighbor algorithm will use the

    highest inverse value to classify new members to the appropriate class. After a new

    member has been classified, he or she is given the learning material for that class. The

    section quiz is availed to the learner and scores determined. The learner is promoted to

    the next level of class when he or she scores above set threshold, otherwise, she or he

    remains in the same group and asked to repeat the class learning process. There will be a

    function which does not employ KNN algorithm to do subsequent classification. When a

    member has been promoted to the subsequent class by the function, the member is given

    a score of the minimum value score of that class and consequently this will affect the

    mean of that class.

    y We shall be proving the concept in a controlled laboratory environment within a day. Weshall set up the system in a laboratory or a controlled room and have learners do the

    learning. The model will then classify the learners according to preset conditions. We

    shall have two sets of learners with one group connecting to connect internet,

    downloading the information to the local database and using it offline to learn and being

    classified. The other group will learn online and be classified. The two groups will then

    be interchanged and the process repeated. We shall determine if the results of both cases

    resemble to prove the concept of intermittent internet connection conditions.

    y We shall design a questionnaire to be answered by students after the test. This will aid ingetting the learners perception of the model. This questionnaire will be in form of hard

    copies and will be provided to the learners. The questions will be designed in a way that

    the learners will be able to state their experience with the model and the appropriateness

    of the model to learning and classification. We shall also be seeking to know if their

    learning performance has improved by using the model.

  • 8/6/2019 Project Research Proposal

    17/23

    Yes

    Yes

    NO

    NO

    Yes

    Start

    Fill initial questionnaire

    Learner meets

    prerequisite?Advise appropriately

    Initial classification

    Give appropriate learning

    material and measure time taken

    Learner read the

    learning material

    and met threshold

    time?

    Give the section quiz and

    determine the score

    Has the learner

    passed this

    section?

    Classify appropriately and

    update the profile Stop

    Give the section

    quiz and

    determine the

    score

    The learner

    passed the quiz?

    The learner

    passed the quiz?

    Give more

    detailed quiz and

    determine score

    NO

    NO

    NO

    Yes

    Yes

    Process Flow Chart

  • 8/6/2019 Project Research Proposal

    18/23

    Proposedsolution

    INTERNET

    Server with Master model

    Laptop with Learner model Laptop with Learner model

    LMS provides

    learning interface

    and share database

    with Master API

    LMS provides

    learning interface

    and share databas

    with client API

    LMS provides

    learning interface

    and share database

    with client API

    Both learning and

    classification takes place

    Both learning and classification

    takes place.Online

    Offline

    Online

  • 8/6/2019 Project Research Proposal

    19/23

    Assumptions and Limitations

    y Classification is only based on the scores and initial knowledge of the learner and howclose they are to other learners already classified. Other learners activities and traits are

    not considered.

    y When learners do not choose the option of going direct to answer questions of anysection, it is assumed that he or she has chosen the option of reading the learning

    materials and time for doing so, is calculated and considered during classification.

    y When learners choose the option of reading the section course material and does not meetthe minimum threshold time, it is assumed that he or she has not read and consequently

    given two quizzes. This is also applied to learners who choose to go directly to answering

    quizzes without reading the section course materials.

    y The model handles only one learning course or subject at a time.y The study will be done in a controlled environment like a laboratory with ten (10)

    students for online and offline sessions and success for both categories compared.

    y Any features of a third party LMS used with or integrated with this model are notemphasized.

    y The LMS and the API share the same database and hence learning material.y Quizzes will be based on the learning materials provided except for the first test which

    will be general knowledge of the course and finding out if course prerequisites are met.

    Tools andfacilities.

    We shall use:

    a) Java language (for web application for online and offline models).

  • 8/6/2019 Project Research Proposal

    20/23

    b) MySql database for hosting learning data. The learning data will be calibrated byindexing for easy access and stratification into different levels to be provided to learners.

    c) A computer laboratory (room) with internet connectivity (where internet can also bedisconnected) for testing the model.

    d) Develop questionnaires to be filled by the learners after going through the process.Timeplan:

    Duration Activity

    October 2010-November 2010 Project proposal, Model research and design.

    December 2010-January 2011 Model development & testing, Data gathering,

    Presentation of the model.

    February 2011-March 2011 User Interface development and testing it with

    the model.

    Report writing and model documentation.

    Project budget:

    Item Cost (Kshs)

    Laptop 70000

    Java Software Free

    MySql Free

    Printing 5000

  • 8/6/2019 Project Research Proposal

    21/23

    Travelling 10000

    Total 85000

  • 8/6/2019 Project Research Proposal

    22/23

    References:

    1. Atolagbe, T. E-Learning: the use of Components Technologies and artificial Intelligencefor Management and Delivery of instruction.Proceedings of 24

    thInternational

    Conference. Information Technology Interfaces. June 24-27, 2002. Cavtat, Croatia. Pp.

    121-128.

    2. Blochl, M. Rumetshofer,H. and Wob, W. Individualized E-Learning Systems Enabledby a Semantically Determined Adaptation of Learning Fragments.Proceedings of14

    th

    International Workshop on Database and Expert Systems Applications (DEXA03). IEEE

    Computer Society 2003.

    3. Chen, Chih-Ming, Hahn-Ming Lee and Yu-Hui Chen. Personalized E-Learning systemusing item Response Theory, Computers & Education, In press, Corrected Proof,

    Available online, 31 March 2004.

    4. Mungunsukh, H. and Cheng, Z. An agent based programming language learning supportsystem. Proceedings of the International Conference on Computers in Education. IEEE

    Computer Society.2002.

    5. Srihivok, A. and A. Intrapairote. A Conceptual Framework for e-learning in the TertiaryEducation in Thailand, Report of the National CouncilResearch of Thailand, Aprll,

    2003.

    6. Zaiane, O.R. Building a recommender agent for e-Learning Systems.Proceedings ofthe

    International Conference on Computers in Education.IEEE Computer Society. 2002.

    7. http://msdn.microsoft.com/en-us/library/1ez7dh12.aspx (accessed on 1st October 2010)8. http://en.wikipedia.org/wiki/Application_programming_interface(accessed on 1st October

    2010)

  • 8/6/2019 Project Research Proposal

    23/23

    9. http://www.codeproject.com/KB/recipes/K_nearest_neighbor.aspx(accessed on 1stOctober 2010)