Withdrawal prediction using the blackboard learning management system through SOM

This paper focuses on the prediction of university student withdrawal prior to completion of their degrees. Unlike traditional dropping-out prediction models, which use demographic attributes and historical records about dropped-out students, the model presented in this article needs only information collected from the blackboard learning management system (BLMS). The indicators used for prediction include student participation in their units and any existing grades from their units. The unsupervised algorithm, Self-Organizing Map (SOM), is used in this prediction model instead of commonly used supervised algorithms. In order to boost student retention rates, a prediction results review model can be added to the existing BLMS. Through analyzing the results generated by the prediction model, mentors can easily find out how the mentees are progressing with their studies, or the mentors may become alarmed and pay more attention to the mentees before their withdrawal.