Predicting At-Risk Students Using the Deep Learning BLSTM Approach

Abstract

Recently, the high usage of online learning platforms by schools and universities has been correlated with an increasing incompletion rate of online courses. Predicting students' academic performance helps the lecturer provide timely intervention and prevent dropping out of classes. This study focuses on applying Deep Learning algorithms to model the learning behaviors of students in a Virtual Learning Environment, predict their performance, and prevent students at-risk from failure. The proposed model is implemented using the Bidirectional Long-Short Term Memory algorithm (BLSTM). Applied to the Open University Learning Analytics Dataset (OULAD), the BLSTM model has achieved relevant results compared to previous approaches namely a cross-validation accuracy rate of 97%.