A set of data mining models to classify credit cardholder behavior

Abstract

In this paper, we present a set of classification models by using multiple criteria linear programming (MCLP) to discover the various behaviors of credit cardholders. In credit card portfolio management, predicting the cardholder's spending behavior is a key to reduce the risk of bankruptcy. Given a set of predicting variables (attributes) that describes all possible aspects of credit cardholders, we first present a set of general classification models that can theoretically handle any size of multiple-group cardholders' behavior problems. Then, we implement the algorithm of the classification models by using SAS and Linux platforms. Finally, we test the models on a special case where the cardholders' behaviors are predefined as five classes: (i) bankrupt charge-off; (ii) non-bankrupt charge-off; (iii) delinquent; (iv) current and (v) outstanding on a real-life credit card data warehouse. As a part of the performance analysis, a data testing comparison between the MCLP and induction decision tree approaches is demonstrated. These findings suggest that the MCLP-data mining techniques have a great potential in discovering knowledge patterns from a large-scale real-life database or data warehouse. © Springer-Verlag Berlin Heidelberg 2003.