Author:

Abstract:

First-year student success in Engineering Bachelor programs is well-studied. Both traditional statistical modelling and machine learning approaches have been used to study what makes students successful. While statistical modelling helps to obtain population-wide patterns, they often fail to create accurate predictions for individual students. Predictive machine learning algorithms can create accurate predictions but often fail to create interpretable insights. This paper compares a statistical modelling and machine learning approach for predicting first-year student success. The case study focuses on first-year Bachelor of Engineering Science students from KU Leuven between 2015-2017 and relates first-semester academic achievement to prior education, learning and study strategies, effort level, and preference for time pressure.