Student Exam Performance Prediction
Predicted student failure risk by analyzing 20+ features including attendance patterns, assignment completion, and grade trends, achieved 93% accuracy and 96% recall, enabling timely advisor intervention. Handled severe class imbalance (75/25 split) by rebalancing training data and adjusting model thresholds to catch at-risk students the model was initially ignoring. Segmented students into 5 behavioral clusters revealing patterns like attendance decline and assignment gaps that preceded failure but werent captured in grade reports.