Student Academic Risk Classification Using Machine Learning: A Comparative Analysis of Predictive Models

Abstract

There are systems in place to identify at-risk students in educational facilities, however early identification allows for more effective support to be allocated. This research presents a framework to identify students at risk in a tertiary institution using contextual, behavioral and academic data. The framework uses a supervised machine learning approach to classify students as Safe, At-Risk or High-Risk. Eight different classification models were trained and evaluated using five-fold cross-validation. The classification performance of the models was analyzed using accuracy, precision, recall and the F1-score. Discrimination analysis was conducted to evaluate the difference in classification. Of the eight classification models, the support vector classifier (SVC) and Logistic Regression model outperformed the others with the SVC achieving 93.63% accuracy. The classification models demonstrated a preference to overfit on the training data. A feature importance analysis was performed to identify the most influential features to the classification models. Some of the more influential features to the classification of risk included students’ previous academic performance and attendance. Based on the results, support systems to identify at-risk students academically can be designed and implemented.