Credit Risk Prediction in Vietnamese Commercial Banks With an Explainable AI Framework Using XGBoost

Accurately predicting credit risk is vital for banks to avoid bankruptcy. Furthermore, there is tremendous pressure to successfully manage credit risk while meeting changing consumer expectations and regulatory norms, yet there is no practical framework in Vietnam for applying advanced machine learning (ML) models in this domain. Traditional models lack the predictive power of newer techniques, but the “black box” nature of ML models makes them difficult for banks to implement. This study develops and evaluates an XGBoost model for credit risk prediction using a dataset from Vietnamese commercial banks, while also designing an explainable AI (XAI) framework to make the model interpretable. By comparing traditional econometric approaches with ML techniques, we aim to create a more effective model for predicting credit risk, using Non-Performing Loans (NPL), Return on Equity (ROE), and Return on Assets (ROA) as key indicators.