Generalized Linear Model for Predicting the Credit Card Default Payment Risk

Abstract

Predicting the credit card default is important to the bank and other lenders. The credit default risk directly affects the interest charged to the borrower and the business decision of the lenders. However, very little research about this problem used the Generalized Linear Model (GLM). In this paper, we apply the GLM to predict the risk of the credit card default payment and compare it with a decision tree, a random forest algorithm. The AUC, advantages, and disadvantages of each of the three algorithms are discussed. We explain why the GLM is a better algorithm than the other two algorithms owing to its high accuracy, easy interpretability, and implementation.