Developing a model to predict defaulting credit card clients using business intelligence techniques

Abstract

The number of personal credit defaults has increased rapidly with the increased popularity of issuing credit cards by banks as a service to customers. Therefore, the risk of defaulters has increased, hence, improving models’ performance to predict defaulting credit card clients is receiving more attention recently. This work aims to develop a model for predicting credit card defaulters by examines the significance of attributes. The results improved by using features selection techniques including change on error approach, filtering infogain attributes, Gain ratio, and Gini algorithms. In conclusion, this work proposed a model that provides 12 attributes as the most important subset features to assessing the risk of defaulters. The Multilayer perceptron algorithm found to give the best performance among other algorithms tested.