Student Risk Assessment: Predicting Undergraduate Student Graduation Probability Using Logistic Regression, SVM, and ANN

Abstract

Understanding how different factors affect the performance of a student in the university setting is important in policy making and providing a better environment for learning. Existing studies on student graduation rates typically employ the use of machine learning methods to correlate a student's profile and their chances of graduation. Building on the success of these methods for Western institutions, we used Logistic Regression, Support Vector Machines, and Neural Networks to build models that use available student data to predict their graduation chances. The results show that all three models are good at predicting graduation outcome, with the logistic regression model yielding slightly higher scores in classification accuracy (80.67 %) and class separation (ROC-AUC score of 83.02%). We also found that including as little as four post-matriculation factors increases the model performances significantly. Hence, the models can be used to perform student risk assessment and develop plans to increase a student's chances of graduation.