Classification of Credit Card Delinquent Customers Using Random Forest Algorithm

Abstract

Customers receive a little plastic payment card called a credit card. It allows the holder access to the market to purchase goods and services mostly on the promise of payment from the holder. The card issuer sets up a revolving account and offers the cardholder a credit line from which they can borrow money for either personal use or to pay for goods or services. To control risks and keep a strong credit portfolio, financial institutions must foresee credit card defaults. In this study, we investigate how to categorize clients who have not paid their credit card balances using the random forest algorithm. We examine the effectiveness of random forest in precisely detecting probable defaulters using a dataset comprising credit card payment history. The study highlights the benefits and drawbacks of the random forest technique through data preprocessing, model application, and performance evaluation. For credit card issuers to make wise judgments and reduce potential financial losses, the findings offer insightful information. The experiment in this study uses real-time data collection obtained from a third-party agency and a random forest technique with tenfold cross-validation. The findings demonstrated that random forest is the best method for categorizing credit card delinquent clients. The classifier outperforms itself in predicting the most likely defaulters with an accuracy of 97.73%, while the classifier’s overall accuracy is 84.703%, according to the experimental data.