Performance Comparison of Convolutional andMulticlass Neural Network for Learning Style Detectionfrom Facial Images

EAI Endorsed Transactions on Scalable Information Systems

F.L. Gambo¹  A.A. Abdullahi¹  G.M. Wajiga²  E.J. Garba²  D.B. Bisandu³  L. Shuib⁴ 

[1] Department of Computer Science, Federal University Dutse, Jigawa, Nigeria;Department of Computer Science, Moddibo, Adama University of Technology, Yola, Adamawa, Nigeria;Department of Computer Science, University of Jos, Nigeria;Department of Information Systems, Faculty of Computer Science and Information Technology, Universiti Malaya, Malaysia;

关键词: learning style;    artificial neural network;    facial images;    vark learning-style model;    deep learning;   

DOI  :  10.4108/eai.20-10-2021.171549

来源: DOAJ

【 摘 要 】

Improvingtheaccuracyoflearningstyledetectionmodelsisaprimaryconcernintheareaofautomaticdetectionof learningstyle,whichcanbeachievedeitherthrough,attribute/featureselectionorclassificationalgorithm.However, the roleoffacialexpressioninimprovingaccuracyhasnotbeenfullyexploredintheresearchdomain.Ontheotherhand, deep learning solutions have become a new approach for solving complex problems using Deep Neural networks (DNNs); theseDNNshavedeeparchitecturesthatarecapableofdecomposingproblemsintomultipleprocessinglayers,enabling and devising multiple mapping of complex problems functions. In this paper, we investigate and compare the performance ofConvolutionalNeuralNetwork(CNN)andMultiClassNeuralNetwork(MCNN)forclassificationoflearnersinto VARK learning-style dimensions (i.e Visual, Aural, Reading Kinaesthetic, including Neutral class) based on facial images. TheperformancesofthetwonetworkswereevaluatedandcomparedusingsquaremeanerrorMSEfortrainingand accuracymetricfortesting.TheresultsshowthatMCNNoffersbetterandrobustclassificationperformanceofVARK learningstylebasedonfacialimages.Finally,thispaperhasdemonstratedapotentialofanewmethodforautomatic classificationofVARKLSbasedonFacialExpressions(FEs).Basedontheexperimentalresultsofthemodels, thisapproachcanbenefitbothresearchersandusersofadaptivee-learningsystemstouncoverthepotentialofusingFEsas identifier learning styles for recommendations and personalization of learning environments.

【 授权许可】

Unknown