Automatic identification of learning styles and working memory capacity from student behaviors using computational intelligence algorithms

Abstract

By identifying students’ learning styles and working memory capacity (WMC) personalized scaffolding techniques can be used, either by teachers or adaptive systems to offer students individual recommendations of learning activities. Such personalization has been shown to have a positive effect on learning outcomes. Traditionally, learning styles and WMC have been identified by dedicated test. However, these tests have certain drawbacks (e.g., students have to spend additional time on them, etc.). Therefore, recent research aims at automatically identifying learning styles and WMC from students’ behavior in learning systems. This thesis presents an investigation into using different computational intelligence algorithms to build automatic approaches to more precisely identify learning styles and WMC. An evaluation of these approaches using real student data shows that most improve precision over existing leading approaches. However the best result for learning styles was a hybrid architecture improving precision styles to 80.4% and an evolving artificial neural network improving precision for WMC to 88.0%. By increasing the precision of learning styles and WMC identification, more accurate interventions can be made to better support students while learning.