Predictive Risk Analysis For Loan Repayment of Credit Card Clients

Defaults on credit card loans lead to a lot of consumer debt and other major problems for banks and financial institutions. This paper compares five different data mining techniques to differentiate between those credit card customers who default on their credit card payments and those who don't. Some techniques use classifiers to identify whether a customer will default on his or her payment such as multi-layer perceptron classifier, while other techniques calculate the probability of defaulting such as logistic regression. Among the classifier techniques used, MLP classifier gives us the best accuracy of identifying the defaulters and Logistic Regression helps us identify some significant factors which have a direct impact on predicting whether a customer will be able to make his or her payment.