The detection of professional fraud in automobile insurance using social network analysis

Abstract: The Automobile Insurance Fraud is one of the main challenges for insurance companies. This form of fraud is performed either opportunistic or professional occurring through group cooperation that leads to greater financial losses, while most presented methods thus far are unsuited for flagging these groups. The article has put forward a new approach for identification, representation, and analysis of organized fraudulent groups in automobile insurance through focusing on structural aspects of networks, and cycles in particular, that demonstrate the occurrence of potential fraud. Suspicious groups have been detected by applying cycle detection algorithms (using both DFS, BFS trees), afterward, the probability of being fraudulent for suspicious components were investigated to reveal fraudulent groups with the maximum likelihood, and their reviews were prioritized. The actual data of Iran Insurance Company is used for evaluating the provided approach. As a result, the detection of cycles is not only more efficient, accurate, but also less time-consuming in comparison with previous methods for finding such groups.