Research On The Classification Of Credit Card Overdraft Repayment Customers

With the development of credit economy,the credit card market in China has entered a period of rapid development.With the increase of cardholders and the increase of card issuance,the accompanying credit risks have increased.While expanding the credit card business at the present stage,banks should pay more attention to the bad overdraft risk brought by credit card customers,increase the intensity of examination and reduce unnecessary economic losses.Based on the data of existing credit card customers of commercial banks,this paper analyzes the overdraft repayment situation of credit card customers through overdraft repayment behavior of overdraft customers.Using descriptive statistics,the important customer indicators that have more influence on overdraft repayment behavior are intuitively analyzed in the form of charts.Then according to the five-level classification criteria of credit card issued by the central bank,the customer status variables were divided into five levels: normal,concern,secondary,suspicious and loss,and the five-level classification data were studied using data preprocessing,data visualization,multi-category modeling and prediction.Finally,based on the practical needs of banks,the classification of customers with bad credit card overdraft are explored.By using machine learning methods such as XGBoost and lightGBM,combined with unbalanced data processing and cost sensitive AdaCost algorithm,a more effective classifier is found to identify small classes of unbalanced data,so as to accurately predict customers who are overdue for a long time and fail to make normal payments.