An algorithm based on fuzzy ordinal classification to predict students’ academic performance

Abstract

Predicting students’ performance in distance courses is a very relevant task to help teachers identify students who need reinforcement or extension activities. Nevertheless, identifying the student’s progress is highly complicated due to the large number of students and the lack of direct interaction. Artificial intelligence algorithms contribute to overcoming this problem by automatically analyzing the features and interactions of each student with the e-learning platform. The main limitations of the previous proposals are that they do not consider a ranking between the different marks obtained by students and the most accurate models are usually black boxes without comprehensibility. This paper proposes to use an optimized ordinal classification algorithm, FlexNSLVOrd, that performs a prediction of student’s performance in four ranking classes (Withdrawn < Fail < Pass < Distinction) by generating highly understandable models. The experimental study uses the OULA dataset and compares 10 state-of-the-art methods on 7 different courses and 3 classical classification metrics. The results, validated with statistical analysis, show that FlexNSLVOrd has higher performance than the other models and achieves significant differences with the rest of the proposals. In addition, the interpretability of FlexNSLVOrd is compared with other rule-based models, and simpler and more representative rules are obtained.