Predictive Model Based on Machine Learning for the Reduction of Student Desertion in Private Universities in Peru: The Case of Universidad Privada San Juan Bautista

Abstract

This study addresses dropout issues in private university institutions in Peru by developing a predictive model using ML techniques. It discusses the relationship between Artificial Intelligence (AI) and Machine Learning, emphasizing the latter's role in enabling computer systems to learn and improve from data. The study explores various reasons for dropout, categorizing forms such as complete, partial, early, and late dropouts, emphasizing the need to address this issue due to its impact on students' academic and professional progress. Using a descriptive and explanatory method, the research analyzes 30 cases from a specific private university institution in Peru. It describes the variables, dimensions, and indicators of the predictive model, highlighting personal factors' substantial impact on the model's predictive capacity. Results reveal significant relationships between variables, with the CatBoostClassifier achieving 78.62% accuracy in early dropout detection. The study underscores the importance of considering personal aspects in preventive and support strategies and presents the predictive model as a valuable tool for addressing student dropout in the Peruvian university context.