Risk Management System for Loan Default Prediction in Banking Sector

Abstract

The paper successfully demonstrates banking loan default risk management through an approach that utilizes XG Boost algorithm and SHAP tool for SHapley Additive exPlanations to provide interpretability to modeling. XGBoost serves as the gradient boosting technique selection because it provides optimal performance with unbalanced data while revealing nonlinear features in borrower information. The high accuracy is achieved by analyzing historical loan data, including both demographic data and credit historical and financial behavioral information. SHAP enables better financial transparency in decision-making procedures by displaying feature contributions that help establish trust and fulfill regulatory standards. Research verification shows that united methods enhance predictive accuracy and create essential risk decisions that specialists need. The system enables contemporary banks to handle speedily approved loans and utilizes the system by developing operational risk mitigation strategies appropriate for present-day banking operations.