To help identify students at risk of not meeting this threshold, our projects focused on using machine learning to improve the accuracy of BSA outcome predictions.

The students explored various machine learning techniques and data strategies to make predictions more reliable and informative. Traditional prediction methods focus primarily on credits earned in the first two academic blocks, but these projects introduced additional factors—such as deviations from average grades, exam attendance, and resit counts—to enhance prediction accuracy. One project specifically used the Synthetic Minority Over-sampling Technique (SMOTE) to address data imbalance issues, as students often cluster in certain credit ranges, which can skew model performance. By using SMOTE and testing 36 different models, including neural networks and logistic regression, the students were able to improve the accuracy of predictions for students with specific credit achievements. The projects are done to help identify at-risk students earlier and allow the university to offer timely academic support.