nurul ilahi husnah

Built end-to-end credit scoring model (AUC 0.7679, KS 0.4026) meeting financial standards. Implemented Reject Inference via Label Propagation to correct selection bias. Developed anti-leakage pipeline excluding post-decision data prior to training. Used XGBoost + SHAP as primary model with Logistic Regression baseline for regulatory compliance, performing threshold tuning to minimize False Rejection Rate.

Developed an end-to-end Credit Risk Prediction model using machine learning algorithms to evaluate borrower default risks for financial institutions. Conducted EDA, data preprocessing, feature engineering, and class imbalance handling. Evaluated models using AUC-ROC and Precision-Recall metrics to optimize credit approval decisioning and minimize financial losses for lending operations.