Balancing Accuracy and Transparency in Early Alert Identification of Students at Risk

One of the challenges in implementing early alert systems to identify students at risk of failure or withdrawal is striking a balance between accuracy and transparency, as there are clear benefits to being able to communicate the reason why a student has been identified. An important predictor of future academic success is past performance, which is generally not available for commencing students. Here, we present a work-in-progress in which the full predictive power of an ensemble-based machine learning approach is employed to make predictions for commencing students, while for ongoing students a simple logistic regression method is used.