Credit Card Default Prediction as a Classification Problem

Nowadays, the use of credit card becomes an integral part of modern economies. Still, predicting credit card defaulters is considered as the most important. So, its assessment becomes a crucial task. In this context, a few Data mining and intelligent artificial techniques were used for extracting meaningful patterns from a given dataset. In this study, we consider credit card risk assessment as a classification problem based on genetic programming (GP) algorithm, where the goal is to maximize the accuracy of the generated model. We evaluate our proposal using customers default payments dataset of Taiwan, and, we compared it with some existing works. The performance of our finding leads to the assumption that GP is able to generate an effective assessment model based on IF-THEN rules. The result confirms the efficiency of our algorithm with an average of more than 86% of precision, recall, and accuracy.