Research On Credit Card Delinquent Behavior Based On Random Forest Classification

Since the reform and opening up,China's personal financial services industry has developed rapidly.Among others,the credit card is one of the fastest growing businesses due to its non-guarantee,overdraft consumption and cash advance.Although China's credit card industry lagged behind the developed countries in Europe and the United States,under the continuous transformation of the concept of national promotion and people's consumption,as of the end of the third quarter of2017,the cumulative issuance of credit cards has reached 552 million and bank card credit The total reached 11.91 trillion yuan.Nowadays,credit cards have irreplaceable status in various financial service products of China's commercial banks.Due to the unsecured and unsecured credit card features,coupled with the early development of major commercial banks in order to seize market share,reducing the threshold for the application of credit cards,making the market a large number of low-quality cardholders,these low-quality Cardholder's repayment ability is not up to standard,over a long period of time there has been a large number of unhealthy arrears,eventually making the credit card market there is a huge risk.Therefore,researching the influencing factors of credit card delinquent behavior and raising the level of credit card risk control are of great significance for promoting the healthy development of credit card market and providing a stable economic environment for our country.In the past,in domestic analysis of factors affecting credit card overdue repayment,most of them were based on the establishment of a Logistic model of personal characteristics.In this paper,combined with the characteristics of the dataused,the explanatory variables are categorical variables,and the explanatory variables include multiple classification variables..This article combines theoretical analysis and empirical analysis for analysis.Among others,the theoretical analysis mainly analyzes the development of China's credit card market,the causes,types,management and prevention of credit risk,as well as the commonly used methods to study credit risk.The empirical analysis studies several methods of categorizing variables in statistical modeling and verification of data with qualitative variables.Based on the data related to the delinquent credit card users of a bank,stepwise Logistic regression and Adaboost and stochastic forest classification models were established to compare and analyze the main factors affecting the credit card overdue behavior.Finally,the three misclassified cross-validation cross-validation of the three methods were compared,the three methods were compared,and the advantages and disadvantages of each method were analyzed.Most of the service products of commercial banks are of the same quality.The characteristics of delinquent related credit card users of a bank reflect the commonalities of the delinquent behavior of most commercial bank credit cards.Through the analysis of variable classification model,it is found that the history delinquent behavior,the credit card utilization rate and the credit limit have a significant effect on the credit card delinquent behavior,and the impact of the housing loan,gender and account opening behavior on the credit card delinquent behavior is not significant.In addition,by comparing the classification methods,it is found that random forest classification has good extrapolation.Through the analysis and research in this paper,combined with the actual situation in our country,it is suggested to improve the relevant policies of credit information in China;improve the credit risk assessment system;and accelerate the establishment of personal credit information system in our country.