A Predictive Model for Dropout Risk in a Computer Science Education Program

Abstract

The ever-growing demand for computing professionals requires the effective management of educational resources. With the increasing importance of computer science education programs in Brazil, identifying potential dropout students has become crucial for educational institutions. However, predicting which students are likely to drop out poses a significant challenge, especially in non-metropolitan areas. To address this issue in the Computer Science Education program of the Federal University of Paraná (Brazil), we propose an approach that leverages machine learning to analyze different features associated with the student's academic performance and detect possible dropouts. We compare the performance of 15 machine learning algorithms in predicting student dropouts, additionally identifying the most influential variables contributing to this situation. To evaluate the effectiveness of our approach, we conduct experiments using real data collected from the computer science education program. The results demonstrate the efficacy of our approach in identifying students at risk of dropping out.