Predicting academic success in Belgium and France Comparison and integration of variables related to student behavior

TL;DR

This study develops prediction models to identify at-risk first-year university students in Belgium and France, highlighting the importance of combining behavioral, personal, and perceptual variables for accurate success prediction.

Contribution

It introduces a comprehensive approach integrating multiple variable types to improve early prediction of academic success among university students.

Findings

Behavioral variables alone are insufficient for accurate predictions.

Adding personal history and perceptions significantly improves model quality.

Early identification of at-risk students can be enhanced with diverse data sources.

Abstract

Having observed low success rates among first-year university students in both Belgium and France, we develop prediction models in this paper in order to identify, at the earliest possible stage, those students who are at risk of failing at the end of the academic year. We applied different data mining techniques to predict the students' academic success. We find that it is very difficult to predict success by only considering the variables related to behavior during classes, and that it is necessary to add variables related to personal history, involvement in and behavior during their studies, and perceptions of academic life, to obtain good-quality results.