Visualizing Credit Default Risk: Insights Through Machine Learning

We apply machine learning to predict credit default risk, aiming to enhance the accuracy of financial risk assessment models and provide valuable insights for risk management. Post-pandemic economic challenges and rising loan defaults have caused significant losses for financial institutions. Existing models fail to deliver sufficient accuracy, and we address this gap by leveraging advanced machine-learning techniques. We design a machine-learning pipeline to analyze and predict credit default risk. We test models like XGBoost, LightGBM, Random Forest, and Deep Neural Networks. We improve performance through feature engineering and optimize models using random search. By combining XGBoost, LightGBM, and Random Forest into an ensemble, we achieve further enhancements. Our final model reaches an AUC of 0.784, proving its effectiveness and laying the groundwork for future advancements in financial risk management.