Adaptive Type-2 Fuzzy Logic Based System for Fraud Detection in Financial Applications

Recent technologies facilitated the use of online transactions and e-commerce as well as simplifying the payment process. In contrast, new fraud patterns are emerging which have many methods, styles and types. The fraud crimes cost financial institutions hundreds of millions of dollars annually which affects the institution financial situation and the customers' confidence. On the other hand, the detection of fraud is very complicated process since the legitimate and fraudulent transactions are similar and it is difficult to differentiate between them as the fraud style is not always same. Hence, there is a need for novel techniques which can detect fraud and adapt to their changing patterns. In this paper, we present our work which is still in progress, to develop intelligent adaptive type-2 fuzzy logic based systems which can detect fraud in financial system. The proposed system is a transparent approach which can learn from data models that could be easily read, analyzed and interpreted by the fraud experts. In addition, the proposed system is adaptive, and it updates itself regularly with patterns which can contain new fraud behaviors. We will focus mainly on credit card fraud and then try to generalize the work to other kinds of fraud.