FUZZGY: A hybrid model for credit card fraud detection

In this paper, we propose a model to detect credit card frauds that can occur in various payment channels. This model is designed to determine the suspicious degree of fraudulent merchant transactions. In our model, fuzzy expert system provides the anomaly degree which clarifies how the new transaction is abnormal in comparison with historical pattern of merchant transactions. The fraud tendency weight is obtained by employing Fogg behavioral analysis. Therefore, this model is called FUZZGY. The fuzzy expert system applies fuzzy rules which effectively use merchant historical activities. These rules are applied to identify any logical contradiction between merchant current activities with trend of her historical ones. The Fuzzy rules are preferred rather than crisp ones because the ambiguous concept of suspicious activity should be considered. Moreover, analyzing the merchant behavior provides valuable information. In fact, it is necessary to evaluate her tendency to make any kind of fraudulent activities. Therefore, Fogg behavioral model is used to describe merchant behavior in two different but related dimensions: motivation and ability to make a fraud. With this model, fraud tendency weight is calculated for each merchant. At the end, suspicious degree of incoming transactions is computed by the FUZZGY model. With successful results which were achieved by testing the FUZZGY in lottery fraud case study, the FUZZGY can be considered as a useful model for credit card fraud detection.