Date Added: Aug 2010
The detection of fraudulent transactions in credit card world is an important application of classification techniques. As human behavior is unpredictable classifying any transaction either as fraud or non-fraud is not acceptable. In all of the previous studies, the transactions were classified in only two levels either fraud or legitimate. An approach where the credit card transactions can be classified in various fraud levels depending on different fraudulent situations mined from the historical behavior of the customers is proposed and implemented.