A Random Forest–Based Framework for Multiclass Student Outcome Prediction and Probabilistic Dropout Risk Stratification

Early identification of students at risk of discontinuing their studies requires integrated analysis of academic and student-related characteristics. This research introduces the concept of AI-Based Student Dropout Risk and Intervention System (ADSIRS) that is based on a multi-class classification of the outcome, risk stratification, intervention suggestion, and role-based surveillance in a single architecture. The pre-processed data about students are applied for Machine Learning (ML)-based classification, and the obtained classifications are further mapped to low, medium, and high-risk categories in order to prioritize the intervention process. Testing showed that the accuracy of ADSIRS was 92.0%, and its multiclass ROC-AUC value was 0.94. This system uses a Python-powered web application that incorporates MySQL to link prediction results with role-based dashboards and intervention alerts. This approach creates an effective computing workflow from the processing of student data and outcomes prediction to risk assessment and interventions, which can be used as a method for student retention management.