An Approach for Non-deterministic and Automatic Detection of Learning Styles with Deep Belief Net

Abstract

AbstractIn spite of the remarkable development of Learning Management Systems, the observation state that these systems don’t fit student learning preferences. To reduce the dropout of student in online learning platforms, learning should be personalized. The migration from classical learning to personalized learning is one way to increase learners’ abilities and improve their learning skills or preferences. This paper defines an approach for developing models of automatic and non-deterministic learning style detection, based on the traces of learner activity in adaptive learning systems. The approach interest lies, on one hand, in its automatic character of the detection of learning traces based on the interactions of learners’ activity in the system instead of a questionnaire used to initialize student Learning styles. On the other hand, in its non-deterministic nature, not like deterministic approaches that do not take into account the uncertainty associated with learning traces, it also uses literature based approached to have an approximation of computed learning styles. Detection is automatic in that unlike traditional methods, it does not question learners to collect information about their learning styles. It is non-deterministic in that it considers the stochastic nature of the learning traces. In the developed approach, DBN-LIS, a generative algorithm, is used to analyse learning traces on LMS, in order to eliminate their inherent stochastic character. The results of this approach are effective, using, unlabelled Moodle and an expert’s labelled dataset, were we reached 91.21% of right detections.KeywordsAdaptive intelligent learning systemsPersonalized learningLearning styleDeep Belief NetFeature selection