May Artificial Intelligence support actions against school dropout?

Abstract

School dropout is a world-level concern due to the negative consequences that it brings to society, so it is important to investigate it to understand and act to mitigate dropout risk. This work proposes the use of Educational Data Mining with Machine Learning to identify variables that are important to characterize the student profile in risk. Support Vector Machine, Gradient Boosting Machine, Random Forest and Ensemble were applied to 1,429 records of undergraduate students in a campus of the IFMG, between 2013 and 2019. The results suggest that Ensemble had the best performance, so it was used to compute the variable importance related to dropout prediction. We used the importance of tracing the student profile of dropout, and proposing a detection and monitoring process to avoid school dropout.

Dropout; Machine Learning; Undergraduate Students