Research On Credit Risk Prediction Of Credit Card Asset Securitization Based On Bp Neural Network Model

As an innovative financial tool,credit card asset securitization has been increasing its issuance scale year by year.Asset securitization products that use credit cards as the underlying assets can increase the capital adequacy ratio and asset liquidity of commercial banks,adding impetus for commercial banks to issue credit card asset securitization.Commercial banks face many risks in the process of credit card asset securitization,among which credit risk is the main risk,so the evaluation of their credit risk is particularly important.The era of artificial intelligence has arrived,and quantification based on artificial intelligence is widely carried out in financial institutions,which provides a certain idea for the prediction of credit risk of credit card ABS products in this article.First of all,this article selects the product with a larger scale of credit card ABS issuance in 2018,namely the "Hexiang 2018-2" credit card ABS product issued by China Merchants Bank as the case study of this article.Based on the introduction of the basic situation of the credit card ABS product in this case,it is determined through analysis that the credit risk of the product is mainly the credit risk of the debtor.Secondly,based on the perspective of predicting the credit risk of credit card ABS based on the BP neural network model,it revealed the influencing factors of credit risk,and selected 23 generated risk factor indicators as the training samples of the BP neural network model.Through the return of the sample data Integrated processing and combing the design ideas of the model to complete the construction of the BP neural network model.In this paper,smooth L1 loss is used as the training loss function,and the prediction performance of the BP neural network model is evaluated through the training and testing of the network.It is obtained from the training process that the BP neural network model designed in this paper is convergent on the data of this case,and then 4 sample data are selected to evaluate the model,analyze the prediction effect of the model through real-time risk and default recovery risk prediction accuracy.The empirical results show that on the trained model,the real-time default rate and default recovery rate prediction accuracy is considerable,and the relative error is relatively low.The BP neural network model constructed in this paper is feasible in the prediction of credit risk of credit card ABS products,and has high practical value for the prevention of credit risk of commercial banks.Finally,combined with the research conclusions,suggestions are put forward to prevent the credit risk of credit card asset securitization.