A Platform for Early Class Dropout Prediction of University Students

Higher education faces significant challenges, with student dropout rates remaining a critical issue. While much research has focused on course-level dropout, less attention has been given to the early identification of at-risk students within individual courses. In this context, this study reviews the state-of-the-art and applies Machine Learning (ML) techniques to predict early dropout of students from real-world data. Key demographic, academic, and behavioral factors were analyzed to identify students at risk of dropping subjects, enabling timely intervention and the implementation of preventive strategies. The analysis revealed that student performance, re-enrollment behavior, and attendance patterns are strongly correlated with the risk of early dropout. Models trained on mid-term and final attendance data before final exams showed the best metrics, highlighting the importance of tracking student engagement throughout the semester. Based on these findings, we developed and implemented the Best Assisting Tutor and INteractive Advisor (BATINA) OnBoard Platform, an advanced platform that integrates predictive models in a practical and user-friendly way. This platform provides comprehensive support to students and faculty, enhancing academic performance and enabling early intervention to improve student retention. This study demonstrates how data-driven solutions can enhance educational support systems and help mitigate student dropout.