Predictive Analytics in Education: Evaluating Machine Learning Methods for Student Dropout Prediction

Abstract

Student dropout remains a pressing concern with significant socio-economic implications. This study utilizes supervised machine learning to forecast potential dropouts by analyzing a diverse array of factors including academic achievements, class attendance, socio-economic backgrounds, and behavioral patterns. These factors are integrated into a comprehensive predictive model that enhances our understanding of student retention and informs the design of targeted interventions. Through a comparative analysis of two prominent algorithms, K-Nearest Neighbors and Naive Bayes, our research assesses the effectiveness of these methods using a detailed dataset. The findings reveal that the Naive Bayes algorithm outperforms K-Nearest Neighbors in predicting student dropouts, offering valuable data for educational practitioners focused on data-driven strategies to enhance student retention. The study advances the application of machine learning in educational settings and contributes practical insights for the development of policies and interventions aimed at reducing dropout rates, thereby enriching the academic discourse and improving educational outcomes.