Method and device for predicting default of credit card user

Abstract

The invention discloses a method and device for predicting credit card user default, and relates to the technical field of bank data processing, and the method comprises the steps: obtaining user information and credit card use information; acquiring a preset first variable set and a data set according to the user information and the credit card use information; repeatedly screening the independent variables in the first independent variable set by adopting a stepwise regression method, and rejecting independent variables without significance; dividing the data set according to a certain proportion to obtain a training set and a test set, and establishing a multivariate logistic regression model according to the screened independent variable set and the training set; inputting the test set into a multivariate logistic regression model to obtain a confusion matrix of overdue defaulted data which is predicted to exceed a first number of days; and calculating according to the confusion matrix to obtain the probability of correctly predicting whether the credit card user is default or not and the probability of the default user, and when the predicted probability meets the prediction requirement, predicting the default condition of the credit card according to the user data by using the multivariate logistic regression model.

Description

technical field [0001] The invention relates to the technical field of bank data processing, in particular to a method and device for predicting credit card user default. Background technique [0002] With the development of the economy, bank credit card business has become popular rapidly and has gradually become an important part of bank revenue. Studying whether credit card users have default conditions will help bank decision makers to better avoid risks and reduce the risk of increased default probability due to credit card users' general reputation. [0003] In the prior art scheme, when using the logistic regression algorithm for default prediction, the independent variables used include the recycling of unsecured loans, the age of the borrower at the time of borrowing, the number of 35-59 days overdue but not bad, debt ratio, monthly Income, number of open credits and loans, number of 90 days past due, number of real estate loans or lines, number of 60-89 days past ...

AI Technical Summary

Problems solved by technology

[0004] However, when predicting default based on the logistic regression model obtained through the above process, descriptive statistics were not made on the data, and the understanding of the data is not enough

The data preprocessing before modeling is not accurate enough, the screening variables are not screened by stepwise regression, and the distribution ratio of training set and test set is not reasonable enough

The model trained in the training set is not applied to the test set, and the accuracy of the model and the accuracy of predicting bad customers are finally calculated

Images