Predicts undergraduate student graduation outcomes using Random Forest, XGBoost, and Logistic Regression with SHAP interpretability and Sankey diagrams to identify at-risk students early.

Comparative analysis of Random Forest, XGBoost, SVM, and Logistic Regression for predicting undergraduate on-time graduation. XGBoost achieved best performance (Acc 82.82%, AUC 0.9122). Dataset: UCI Predict Students' Dropout. Python 3.10 · Scikit-learn 1.3.0 · XGBoost 2.0.0