Credit Default Prediction with Machine Learning: A Comparative Study and Interpretability Insights

This study built an efficient credit default prediction framework based on the LightGBM model, aiming to improve the prediction performance and enhance the interpretability of the model. Through experiments on the public Home Credit Default Risk dataset, this paper comprehensively compared LightGBM with SVM, XGBoost, GBDT, and random forest models, and verified the superiority of LightGBM from multiple indicators such as accuracy, recall, and F1-score. The experimental results show that LightGBM performs best in processing complex high-dimensional data with its efficient feature selection and optimization capabilities. At the same time, combined with SHAP value analysis, it reveals the impact of features on model decisions, providing transparent and reliable decision support for credit risk assessment. This study not only provides an effective technical solution for financial institutions to improve risk management efficiency but also lays the foundation for the application of machine learning algorithms in the financial field. In the future, multimodal data and deep learning technology will be further integrated to build a more powerful prediction model.