Modelo predictivo de alerta temprana para identificar a los estudiantes con alto riesgo de abandonar sus estudios universitarios

Abstract

Abstract: University dropout represents a critical challenge in higher education, with both academic and social implications. This study presents the development of an early warning model to identify students at elevated risk of dropping out, combining machine learning and deep learning approaches. An extensive literature review was conducted to identify factors associated with dropout, including academic, socioeconomic, personal aspects, and more. The data used was sourced from UTPL databases, integrating features such as academic performance, platform usage, among others. Various machine learning algorithms, including tree-based models, ensemble methods, and others, as well as deep neural networks, were assessed to evaluate their performance. The results indicated that deep learning models achieved higher accuracy in prediction. However, machine learning models excelled in assigning risk scores for student dropout, providing incredible useful data for the Higher Education Institutions (HEIs). This model aims to serve as a practical tool for HEIs, facilitating early interventions that promote student retention, thereby contributing to academic success.