Application of Machine Learning Algorithms in Credit Card Default Payment Prediction

Credit card default payment prediction studies are very important for any financial institution dealing with credit cards. The purpose of this work is to evaluate the performance of machine learning methods on credit card default payment prediction using logistic regression, C4.5 decision tree, support vector machines (SVM), naive Bayes, k-nearest neighbors algorithms (k-NN) and ensemble learning methods voting, bagging and boosting. The performance of the algorithms is evaluated through following performance metrics: accuracy, sensitivity and specificity. The best result among all algorithms for overall accuracy rate was achieved by logistic regression model with a rate of 0.820. The best performing model for default credit card customer detection, with success of 71,3% was naive Bayes model. This approach could improve and ease the process of credit card default, and therefore help the banking system in decision making