A study of machine learning based credit card potential default customer identification

Abstract

Fintech is continuously driving the overall upgrade of payment methods. Technologies such as Big Data, the Internet of Things, and Artificial Intelligence continue to be applied in the payment field and significantly impact the payment industry. Based on the fusion of multiple machine-learning models, the problem of identifying potential default credit card customers is investigated in this paper. The customers are mined and classified based on their bill amount, education level, marital status and other characteristic information. The various models predict using the AutoML framework, then fused and optimized by bagging and stacking methods, and the models are evaluated using evaluation metrics such as F1 values. The test results show that the F1 value of the integrated model after multiple stacks reaches 54.3%, which is better than that of a single algorithm.