A Machine Learning Approach to Identifying Students at Risk of Dropout: A Case Study

The increase in students’ dropout rate is a huge concern for institutions of higher learning. In this article, classification techniques are applied to determine students “at-risk” of dropping out of their registered qualifications. Being able to identify such students timeously will be beneficial to both the students and the institutions with which they are registered. This study makes use of Random Forest, Support Vector Machines, Decision Trees, Naive Bayes, K-Nearest Neighbor, and Logistic Regression for classification purposes. The selected algorithms were applied on a dataset of 4419 student records obtained from the institutional database related to Diploma students enrolled in the Faculty of Information, Communication and Technology. The results reveal that the overall accuracy rate of Random Forest (94.14%) was better than the other algorithms in identifying students at risk of dropout.