运用机器学习技术提高学生有效的学习计划(IJMECS-V9-N3-1)

I.J.ModernEducationandComputerScience,2017,3,1-9PublishedOnlineMarch2017inMECS()DOI:10.5815/ijmecs.2017.03.01Copyright©2017MECSI.J.ModernEducationandComputerScience,2017,3,1-9EnhancingEfficientStudyPlanforStudentwithMachineLearningTechniquesNipapornChanamarnDepartmentofComputerScienceandTechnology,FacultyofScience,NaresuanUniversity,Phitsanulok65000,ThailandEmail:[e-mail]ntofComputerScienceandTechnology,FacultyofScience,NaresuanUniversity,Phitsanulok65000,ThailandResearchCenterforAcademicExcellenceinNonlinearAnalysisandOptimization,FacultyofScience,NaresuanUniversity,Phitsanulok65000,ThailandEmail:[e-mail]—Thisresearchaimstoenhancetheachievementofthestudentsontheirstudyplan.Theproblemofthestudentsintheuniversityisthatsomestudentscannotdesigntheefficientstudyplan,andthiscancausethefailureofstudying.MachineLearningtechniquesareverypowerfultechnique,andtheycanbeadoptedtosolvethisproblem.Therefore,wedevelopedourtechniquesandanalyzeddatafrom300samplesbyobtainingtheirgradesofstudentsfromsubjectsinthecurriculumofComputerScience,FacultyofScienceandTechnology,SakonNakhonRajabhatUniversity.Inthisresearch,wedeployedCGPApredictionmodelsandK-meansmodelson3rd-yearand4th-yearstudents.Theresultsoftheexperimentshowhighperformanceofthesemodels.37studentsasrepresentativesampleswereclassifiedfortheirclustersandwerepredictedforCGPA.Aftersampleclassification,samplescaninspectallvectorsintheirclustersasfeasiblestudyplansfornextsemesters.SamplescanselectastudyplanandpredicttoachievetheirdesiredCGPA.TheresultshowsthatthesampleshavesignificantimprovementinCGPAbyapplyingself-adaptivelearningaccordingtoselectedstudyplan.IndexTerms—MachineLearning,Prediction,Clustering,GradeDataPatterns,StudyPlanI.INTRODUCTIONEffectiveundergraduateeducationisakeycomponentthatsupportsthedevelopmentofthenationalhumanresourcesinthecountry.Studyinginundergraduatelevelaimstodeveloppersonnelskillsandknowledgethatarenecessaryforparticularcareers.Therefore,curriculumsareorganizedwithcourseswhicharecarefullydesignedtoensurethatstudentsarereadyfortheirupcomingoccupations.However,someundergraduatestudentshaveafailureinundergraduatestudyingbecausetheyhaveacumulativegradepointaverage(CGPA)lowerthanthedefinedstandard.Thissituationofstudentscanleadtothedelayoftheirgraduationandlossoftheopportunitytofindgoodwork.Manyeducationalinstitutionshavewastedannualbudgetstosolveanincreasingrateofdropoutstudents[1].Academicstafftriestotacklethisproblembyincreasingstudentefficacyandsuggestingaguidelineforbetterstudyperformance.Manyscholarstrytohelptostudentsbyapplyingsciences,includingusingpsychologicalcounselinginstrument[2]andstatisticalsciencetoanalyzelearningassessment[3].EducationDataMiningtechniques(EDM)applymethodsandtechniquesfromstatistics,datamining,andmachinelearningtoanalyzetheeducationdata[4-5].Recently,MachineLearning(ML)emergedandgainpopularitywhichisaverypowerfultechniquefromcomputationalandstatisticalmethodsusedforpredictionsanddatapatterninferences.MLcanbeusedtosolvemanyproblems,suchasproblemsaboutmarketing,banking,medical,industry,agriculture,andpowergeneration[6-10].Furthermore,itwasusedineducationdomainforstudyinginhighereducationinstitution.ResearchersineducationmostlyapplyMLintwotasks.Thefirsttaskisapredictiontaskforstudentperformance,andthesecondtaskisaclassificationtaskwhichclassifiesstudentsintogroupsbasedonthestudents’profiles.Inpredictiontask,severalMLtechniqueswereusedinresearch.Forexample,[11]gradesofstudentsinbasicengineeringsubjectsandbasicmathematicssubjectswereusedwithNeuralNetworktopredictCGPA,and[12]gradesofstudentsinEnglishcoursewereusedwithNeuralNetworktofindCGPA.[13-14]VariousMLtechniqueswerecomparedinCGPAprediction,whichincludeRadialBasisFunctionNetwork,NeuralNetwork,andSupportVectorMachinebyusinggradedatafrombasicsciencesubjects.Inclassificationtask,someresearchersusepredictedCGPAtoclassifystudentsintodifferentclasses.Thefollowingexamples[15-16]appliedNeuralNetworktoclassifygroupsofstudentachievementperformancebyusinggradedataineachsemester.[17]usedDecisionTreetoclassifygradesinbasiccomputerprogrammingsubjectsinto2EnhancingEfficientStudyPlanforStudentwithMachineLearningTechniquesCopyright©2017MECSI.J.ModernEducationandComputerScience,2017,3,1-9achievementclasses.Althoughalldescribedstudiesusesubjectsgradeasinputdata,somestudies[18-22]usebothgradesandstudents’profiles;sex,age,highschool,andparenteducationareusedasinput.Inclassificationtechniquesevaluation,manyMLtechniqueswereapplied[17,20-22].In[23],evaluationwasanalyzedonvarioustechniques,includingDecisionTree,NeuralNetwork,andNaïveBayes.[24]comparedbetweenseveraltechniques,includingDecisionTree,NeuralNetwork,NaïveBayes,SupportVectorMachine,K-NearestNeighbour,andLinearRegressiontofindwhichisthebestMLtechnique.Accordingtoreviewedpaper,themostuseddatasetforpredictionareCGPAandinternalassessmentscore.WhiletheNeuralNetworkandDecisionTreearethetwomethodsthataremostlyusedbytheresearchersforpredictingstudentperformance[