Predicting Students' Academic Performance in Virtual Learning Environment Using Machine Learning

Abstract

The Open University (OU), one of the largest public research universities, provides a wide range of data from its distance learning courses. Hence, the Open University Learning Analytics Dataset (OULAD) allows predicting student academic performance in online learning programs. The dataset consists of demographic features such as gender, disability, education level, and behavioural features, which depict engagement levels of students in courses. This paper predicts student academic performance in online learning programs using machine learning and statistical values. We train multi-class classifiers on the preprocessed dataset after feature selection and removing noisy data. Decision Tree, Random Forest, Gradient Boosting and KNN classifiers are trained on both demographic data alone and including virtual learning environment (VLE) data with it. Each classifier shows greater accuracy with the VLE data included. All classifiers achieve accuracies above 92%, with gradient boosting achieving the maximum accuracy of 97.5%.