Monitoring student performance and identifying academically at-risk learners at an early stage is a major challenge in educational institutions. Delayed identification limits timely academic intervention. The proposed system, Student Performance Prediction and Early Warning System, is a data-driven platform designed to analyze student behavior and predict potential academic risks. The system considers academic and behavioral parameters such as study hours, previous semester scores, sleep duration, weak subjects, attendance, and assignment performance. Enhanced features including attendance percentage, study efficiency, and academic improvement trends are incorporated using feature engineering. Data preprocessing techniques such as missing value handling, encoding, scaling, outlier detection, and SMOTE-based data balancing are applied for improved model performance. Machine learning models including Logistic Regression, Decision Tree, Random Forest, along with advanced models like XGBoost and LightGBM, are used with hyperparameter tuning to classify students into high, medium, or low risk categories. The system consists of student and administrative modules. Predictions can be made using the student module that relates to the performance and customized AI-based study guides per topic. In addition, the administrative module helps monitor the performance of the students through filters, searches, and graphs. This will help identify any at-risk students. It gives the opportunity to take proactive measures and make sound judgments for academic success.
