Remedial actions recommendation via multi-label classification: A course learning improvement method

Abstract

In this paper we present RARS - Remedial Actions Recommender System that is based on a multi-label classification approach to recommend remedial actions to address student performance shortcomings in Learning Outcome Attainment Rates. A dataset of rubric instances is constructed where each instance is characterized by a set of features (e.g. course domain, course level, etc.). Classes labeling the training instances correspond to the remedial actions that have been proposed by instructors and Quality Assurance Experts over several semesters. Experiments carried out on the constructed dataset showed that the use of wrapper multi-label classification approaches as a basis of RARS and especially the classifier chains method with decision trees as a base classifier provides useful remedial actions recommendations.