Predict Student Dropout Rates in Higher Education Using Academic and Non-Academic Factors: A Machine Learning Approach

Abstract

Student dropout in higher education poses a significant challenge to academic institutions worldwide, often leading to reduced institutional performance and compromised student success.This study aims to address this problem by developing a machine learning-based predictive framework that integrates both academic and non-academic factors to identify students at risk of dropping out.Utilizing a Random Forest Classifier trained on publicly available datasets, the model analyzes variables such as GPA, attendance, financial aid status, and extracurricular involvement.The predictive system was tested through a user-friendly Flask web application, enabling both batch and manual predictions with high accuracy.Evaluation metrics, including accuracy score, ROC-AUC, and confusion matrix, confirm the model's reliability and robustness when tested with real-world data.The results demonstrate that combining academic and behavioral indicators enhances the precision of dropout detection and provides valuable insights for designing early intervention strategies.This research contributes to educational data analytics by offering a scalable, interpretable, and actionable tool for improving student retention in higher education.