Predicting Freshmen Students’ Academic Performance and Mental Health in Higher Education Using Machine Learning

Predicting undergraduate university students’ poor academic performance is crucial for reducing its adverse effects on their mental health, enhancing students’ ability to cope with stressful situations, and thereby reducing the harm it causes to their overall well-being. This study examines the factors contributing to poor performance among freshmen students in the Information Systems program, leading to the development of stressful behavior and eventual program dropout. Factors such as students’ gender, nationality, high school type, high school score, English level, performance in mathematics, performance in programming, and grade point average, among others, were investigated using three machine learning algorithms: Extra Trees, Light Gradient Boosting Machine, and Random Forests. The results indicated that the grade point average is the most significant factor affecting students’ mental health, resulting in their leaving the program. Performance in mathematics and gender are the other factors affecting students’ stress behavior. After investigations, our results found that among students who performed poorly, males left the information systems program more than their female counterparts. They mainly transferred to less technology-oriented majors, such as management and law. In addition, it was found that women performed better in mathematics among students leaving the program than men. These results can be a central priority for educators, policymakers, and community members to positively affect students’ academic performance and mental health.