基于证据权重及信息价值筛选的信用卡透支风险检测模型的建构

Credit Card Overdraft Risk Detection Model Construction Based on the Weight of Evidence and Information Value Screening

研究通过kaggle数据挖掘竞赛提供的持卡人特征、信用历史等信息,采用证据权重、分段等方法测算了各变量对是否违约的影响,并使用信息价值筛选出了贷款利用情况、信贷经历、年龄等关键因素,构建的信用卡评分模型(AUC)值为0.85,其检测方法真实性较高,正确率较高。据此银行可从持卡人的信贷经历、贷款利用情况考查持卡人透支信用卡的风险。

We study the cardholder characteristics,credit history and other information provided by Kaggle data mi-ning competition,using the weight of evidence and segmentation methods to measure the impact of each variable on default,and using the value of the information to screen out key factors such as loan utilization,credit history,and the age.The constructed credit card scoring model has an AUC value of 0.85,and the detection method is authentic and correct.The bank can check the cardholders′risk of overdrafting the credit card from their credit experience and loan utilization.