Exploring the Suspicious Behavior in Electronic Banking System

One of the most crucial impediment in bona fide utilization of electronic banking services in Iran is de facto¡ the observed misuses of its functions which may lead to strong sense of uncertainty and insecurity for conducting the financial transactions per se. Hence in this context¡ exploiting the appropriate methods for identifying the suspicious behavior will be deemed as a critical issue for financial institutions and banks¡ in toto. In this paper¡ an attempt is made¡ to discern the phenomenon of fraud in bank cards¡ viz-a-viz exerting de jure¡ the technique of Data Mining¡ thereto. Though it is worth to note that¡ we lack the authentic statistics to specify the exact number of fraudulent incidences in valid bank cards of Iranian banking system¡ but we have hypothesized that the prevalence of fraud in bank cards in this country is growing¡ and we expect that in near future it could become one of the main hindrance of banking system in Iran¡ Ipso facto. Therefore¡ initially we have applied Delphi method to acquire the expert opinions regarding the concerned issue¡ and diagnosing the current state of fraudulence in bank cards¡ via inducing significant changes on data by applying the simulation technique to trace out the fraudulent transactions. Subsequently¡ we have tried to design a model to classify the transactions into two subgroups of genuine and fraudulent ones¡ using the Clustering Method in Data Mining and employing three models of Neural Network¡ Decision Trees and Nearest Neighbor¡ Quid pro qua. The results prima facie¡ indicate that the Decision Trees model¡ among three methods under the review¡ carries out the highest significance with 98 percent of confidence in specification of fraud in Karafarin bank cards¡ Sine qua non.