Research On Credit Card Risk Assessment Based On Machine Learning

China's credit card business has ushered in a fast-rising development stage after a slow market cultivation phase.Major banks are greatly attracted by the high profits of credit card business and the huge market space,and they strongly increase the issuance of credit cards.However,high profits are accompanied by high risks.When faced with tens of thousands or even more users who apply for loans,whether or not to lend will be determined.The traditional model and method alone cannot satisfy the status quo.It requires the use of Internet and computer technologies to solve the problem of forecasting credit risk.This article uses the Give me Credit dataset on Kaggle,the U.S.data competition platform.Using machine learning methods,logistic regression,decision tree,random forest model xgboost four main classification models for analysis,established a credit card risk assessment model to determine whether the applicant would overdue or not repay the loan,achieved a quantitative analysis of personal credit risk.This article starts with the basic concepts and characteristics of credit cards and their risks,and summarizes the research results of credit card risk assessment models.The theory of machine learning,the principle of the algorithm and the evaluation indicators of the model are elaborated in detail,and credit card risk assessment is performed through machine learning.In the specific experiment,this paper uses machine learning technology to model and analyze the credit card users.Among them,the SMOTE method is used to process the data oversampling,outliers,and missing values,and the variables are standardized to make them The numerical range falls within the same range.Finally,logistic regression,decision trees,and random forests and xgboost are used for predictive modeling.The Recall value is used as a specific indicator of performance verification of the model to further optimize the model.Finally,the results show that the results of xgboost,whether in the test set or the training set,are superior to other algorithm models,and further results are obtained.Using the xgboost algorithm in credit card risk assessment can get a good prediction.Compared with other methods in the literature,new ideas are provided for reference and have certain guiding significance.