Temporal Analysis of Active Students in Machine Learning

University dropout represents a growing challenge for institutions, particularly during the first academic year. This thesis investigates whether machine learning models can predict students’ regularity during the initial months of enrollment. This could represent a reliable early warning system to identify at-risk students, enabling timely interventions. The results revealed that meaningful predictive signals emerge as early as January; suggesting that the timing and quantity, of early exams can serve as reliable indicators of later academic regularity. These findings demonstrate the potential of data-driven approaches to support Universities in monitoring student progress more effectively.