An empirical study of feature extraction in the analysis of credit card risk

The default risk is a critical loss to credit card business. Various methods of early detection start with meaningful features to obtain good performance. Lack of detailed data has limited the understanding of consumer behavior and motivation in the use of cards and subsequently in a more complete understanding of default. We discuss several ways to extract all kinds of features to illustrate the indication of them to distinguish the good and bad accounts. With the availability of the real data, a new and more complex picture of cardholders´ behavior has emerged.