Research on Bank Credit Risk Assessment Model based on artificial intelligence algorithm

Abstract

With the rapid development of China's economy and the generation of over-consumption concept, major banks are facing serious credit risk problems therefore, it is especially important to establish a scientific and effective risk assessment model for the healthy development of their banks. Based on artificial intelligence algorithm, this paper constructs an integrated classification model through Stacking, combined with SMOTE (Synthetic Minority Over-Sampling Technique) oversampling method, to analyze and evaluate the risk of customer credit, which can help banks to effectively identify potential credit default customers and reduce the loss of banks. The data in this paper are obtained from LendingClub. Firstly, the Stacking model integration method is used. The results show that integrated model has the characteristics of high accuracy and high robustness with the Stacking model integrated method. The accuracy score of the integrated model is 0.83, and the stability score is 1.00. At the same time, SMOTE oversampling method is used to optimize and improve the problem of unbalanced raw data, which reduces the overfitting phenomenon and enables the model to identify more defaulted customers and improve the prediction effect on a few defaulted customers.