Research on bank credit risk control based on AE-CatBoost

With the continuous progress of society, the scale of credit business is also expanding. This thesis conducts research on machine learning to accurately assess borrowers' credit risks based on a public data set of loan records from a certain credit platform. By combining the deep learning model AutoEncoder and the integrated learning model CatBoost, the AE-CatBoost model was proposed. Various evaluation indicators were compared through experiments and it was found that the AUC value of AE-CatBoost model is about 0.034 higher than that of CatBoost model. The proposal of AE-CatBoost has important practical significance for the development of bank credit risk control.