An Empirical Study On Evaluation Of Credit Card Default Risk For Commercial Banks

Credit card business is a typical weak sensitive asset in the business cycle of "small amount dispersion" and "asset light".With the prosperity of the consumer credit market in recent years,it has experienced considerable development and still has huge potential.However,the credit card business has more or less hidden risks during its establishment and operation,and there are also many deficiencies in the risk measurement and management level of commercial banks in China.If a large amount of credit card customer information recorded on multiple internal system platforms can be effectively utilized,and big data mining methods are used to establish a model learning and iterative architecture,improve the accuracy of risk assessment,strengthen admission interception,early warning control,commercial banks can achieve the goal of controlling default losses,and even significantly reduce default losses.Firstly,this paper summarizes the theoretical basis and research achievements of default risk management and default risk assessment technologies at home and abroad,and designs the process structure of default risk assessment and control.Then,the label set of historical data such as basic information classification,transaction repayment behavior,internal and external credit status of some credit card customers of S commercial bank was adopted.After data cleaning and characteristic variable analysis,the credit card default risk assessment model was established by using Logistic regression,Random Forest and XGBoost algorithms respectively.And combined with the model effect transformation to develop a more intuitive risk score card.By comprehensively comparing the prediction accuracy,default recall rate,AUC and KS of the three models,it is found that XGBoost model has the best comprehensive effect,and XGBoost scorecard has a higher default sample recall rate,while Logistic regression scorecard has a stronger interpretative ability.Therefore,commercial banks can use XGBoost algorithm to establish credit card default risk assessment mechanism,monitor and output default probability with high accuracy,make objective evaluation of customers’ credit situation combined with score cards,and simultaneously use Logistic regression to help master default factors.The above research results can help banks improve the technical level of predicting the default possibility of credit card customers in the future period and effectively quantify risks.Meanwhile,banks should expand and improve the whole-process risk management and risk pricing system based on the default risk assessment module.Finally,this paper puts forward some suggestions on credit card risk management.In summary,this article has laid a rich theoretical foundation for commercial banks to improve their credit card risk management system and risk management value concept,and has certain theoretical and practical significance for improving the measurement and management level of credit card default risk.