CatBoost And Its Application In Personal Loan Credit Evaluation

With the advent of the era of big data,domestic banks and financial institutions have to transform to technological digitization in the face of massive and chaotic data.Personal credit risk assessment has also become a major research focus of major banks and financial institutions.This paper builds a personal credit evaluation model based on the CatBoost algorithm that supports category features.The symmetric decision tree is used as the learner during model training to improve the prediction efficiency of the model.The Ordered Boosting method avoids the problem of prediction bias and improves the accuracy of the model.degree.Empirical research using real loan data from the Lending Club platform,and a comparative analysis with the traditional gradient boosting algorithms GBDT and XGBoost models.The results show that the CatBoost algorithm is superior to the traditional gradient boosting in classification accuracy and category discrimination ability algorithm.Finally,the prediction results of the CatBoost model are explained globally and locally under the SHAP framework.