This paper proposes a HMM (Hidden Markov Model) based fraud detection system for credit card fraud detection. The method works on the statistical behavior of user's transactions. Since the original transactions are not available due to privacy policies of bank we used here synthetically generated data for a credit card user, and then HMM model is trained using different size of sample of generated labeled data we also discuss the performance of the HMM model on this data set in terms of detection accuracy and earliness of fraud detection. The system has been tested on a Pentium 4 PC with 2 GB of RAM, the test program is coded in MATLAB 7.5.
Fraud detection, HMM (Hidden Markov Model)
