CREDIT CARD DEFAULT PREDICTION USING MACHINE LEARNING

Abstract

The credit card industry had been around for decades and is a product of changing consumer habits also increasing the national income. There has been a significant increase in the number of card issuers, issuing banks to transaction volumes. However, with the increase in credit card transactions, the amount due and the arrears rate of credit card loans are also issue that cannot ignore. This issue is crucial for the successful development of the banking industry in the future. The study focused on modeling and predicting an individual's willingness to repay credit card loans. The methods used in this study are machine learning with random forest approach, artificial neural network, support vector machine, logistic regression, and naïve Bayes. There are 11 variables to be analyzed in this study, and the performance of the five methods will be compared to the evaluation of ROC, and AUC. The result of this research as follows. The random forest method is considered the most appropriate for processing the credit card default dataset with AUC 89%. This model can contribute to the settlement of default probabilities and is of great help to the credit card industry. Based on the PDP, managerially it can be determined that for income and credit card limits the range of 7-50 million is more prone to default