StayOnTrack: An Advanced Student Dropout Prediction and Intervention System Using Explainable Machine Learning

Student dropout in higher education remains a critical challenge, leading to significant academic and institutional losses. Traditional monitoring systems are often reactive and fail to identify at-risk students in a timely manner. This paper presents "StayOnTrack", an advanced data-driven decision support system designed to predict student dropout risk and enable targeted interventions. The proposed system integrates multi-dimensional student data, including academic performance, attendance, financial status, and learning management system (LMS) engagement. It employs an ensemble machine learning approach combining CatBoost and deep neural networks to generate a dynamic dropout probability score for each student. To enhance transparency and trust, the system incorporates Explainable Artificial Intelligence (XAI) using SHAP (SHapley Additive exPlanations), providing interpretable insights into the factors influencing each prediction. The solution is implemented as a full-stack web application using modern technologies, ensuring scalability and usability in institutional environments. Experimental results demonstrate efficient performance, with rapid inference times and high system reliability. The proposed approach offers a proactive strategy for improving student retention and supports data-informed decision-making in higher education institutions.