due to extraordinary large amount of information and daily sharp increasing claimant for ui benefits and because of serious constraint of financial barriers, the importance of handling fraud detection in order to discover, control and predict fraudulent claims is inevitable. we use the most appropriate data mining methodology, methods, techniques and tools to extract knowledge or insights from enormous amounts of data to detect white-collar crimes. fraud detection in ui as a category of social benefits is used as the application domain. the problem solving approach is to integrate database, machine learning, neural networks, data visualization, statistics and distributed data mining techniques and tools into the crime detection system. the crime detection method utilized bayesian belief network (bbn), decision tree (dt) and multi layer feed forward neural network (mlnn) learning algorithms; classification and clustering visualization. the most important predictor variables have identified to build a model for fraud pattern recognition. when a claim is processed, it will classify as “normal”, “fraudulent” or “high risk fraudulent”. also as a subsidiary analysis, performance of various data mining techniques have compared with each other in some issues in available insurance data. these findings depict a clear image of rising fraud from a simple claim to a disastrous and disruptive form, and also the results will provide an appropriate opportunity for social insurance authorities to have better conceptual understanding and effective combat against all kind of insurance fraud, just on time.