Credit Card Fraud Detection Using a Neuro-Fuzzy Expert System

In this paper, a two-stage neuro-fuzzy expert system has been proposed for credit card fraud detection. An incoming transaction is initially processed by a pattern-matching system in the first stage. This component comprises of a fuzzy clustering module and an address-matching module, and each of them assigns a score to the transaction based on its extent of deviation. A fuzzy inference system computes a suspicious score by combining these score values and accordingly classifies the transaction as genuine, suspicious, or fraudulent. Once a transaction is detected as suspicious, a neural network trained with history transactions is employed in the second stage to verify whether it was an actual fraudulent action or an occasional deviation by the legitimate user. The effectiveness of the proposed system has been verified by conducting experiments and comparative analysis with other systems.