Research On Post Loan Risk Collection Scoring Model Based On Logistic

In recent years,with the rapid development of personal consumption credit in the field of Internet finance,the issuance of personal credit on the credit platform tends to be fast and simple,but this also increases the risk of overdue in the market.Especially in the Indonesian credit market,the delinquency rate of personal credit has been high,which poses a huge risk for collection.In order to quantify the collection risk of customers,this paper introduces a collection scoring model and adopts a differentiated collection strategy for customers according to the collection scoring.This paper collects raw data from a domestic financial technology company's Indonesia microfinance loan product.Through data preprocessing,exploratory data analysis,variable selection,model development and evaluation process,a WOE-encoded collection rating model is established.Using ROC curve and KS value as the model of evaluation index,the test set evaluation results show that the AUC value is 0.73,KS value is 0.34.In addition,in order to check the robustness of the model,another month's raw data are collected to be substituted into the score card as a new test set,and the evaluation results show that the AUC value is 0.76 and KS value is 0.4.The validity of the model is proved by two tests.Based on the score distribution and default rate of the scoring card,the customers are divided into groups A,B,C,D and E.Group A has the highest score and the lowest risk,while group E has the lowest score and the highest risk.On the basis of understanding the strategies used in the reminder process,the reminder strategies with different intensity are used for groups with different risks.This is expected to save at least 30%of the original human resources,and to identify high-risk users accounting for 10%of the total customers and to prepare for the subsequent overdue risk.The collection score card obtained in this empirical study not only makes a high-precision prediction in the case of fixed time span,but also enables the variables to have a good explanatory ability,and enables the business personnel to judge the customers by understanding the variables,which has high practicability and generalization.