Credit Risk Prediction of Listed Companies Based on Smote-RF-Catboost

Abstract

The credit risk of listed companies is the focus of attention in the field of financial risk, and accurate identification and prediction of credit risk is of great significance to the healthy development of China's financial market and the smooth operation of the financial system. In this paper, a Smote-RF-Catboost method is proposed to study A-share listed companies in China, which selects 23 indicators that can comprehensively reflect the financial situation of enterprises to build an index system, combines Smote (Synthetic Minority Over-Sampling Technique) over-sampling to achieve sample equilibrium, and uses Random Forest feature importance to select features. Finally, the Catboost algorithm is used to predict the credit risk. The empirical results show that the prediction accuracy and AUC value of the test set are above 0.95, and the prediction performance is high, which provides a new way for financial institutions and related practitioners to analyze and predict credit risk.