Study On Medical Insurance Fraud Identification Of The New Rural Cooperative Medical System Based On Outlier Detection

The new rural cooperative medical system is a preferential policies which established by Chinese government. Its aim is reforming the rural health care system, improving the rural quality of medical service and abating the economic problems caused by illness among the rural people. It has great significances for promoting the coordinated development of urban and rural areas and resolving the Three-agricultural problem. Since the implementation of the system, exploring the system operating mode and expending in scale have always been the key work in different regions. However, the risk of funds security often been neglected by people. Because the new rural cooperative medical system has the characteristics of the insurance industry, the existence of moral hazard inevitably caused all kinds of health care fraud and other illegal events. It took serious economic losses and influenced the system stable and sustained long-term development. To deal with the problem, adopting outlier detection technology to identify the medical insurance fraud has great practical significance for helping audit staff complete the examination and supervision.Outlier detection is an important direction of data mining research, this paper first analyzed the concepts of data mining, discussed the data mining process and focused on the outlier detection about the methods of data mining. Aiming at a few kinds of existing methods taken a comprehensive and richly detailed analysis from several aspects. Such as the basic thinking, advantages, disadvantages and scope of application.Secondly, considering the characteristics of the data of the new rural corporative medical system, this paper proposes an improved algorithm of two-stage outlier detection-TSOD, based on the existing methods. It continues the outlier factor concept of the density-based outlier detection. The goal of the algorithm is find out the Top-n outliers in the data set. Combining with clustering analysis technology to reduce the size of the data set, and using a pruning rules in the process of search K-Nearest Neighbors, improved the efficiency of the algorithm.Finally, according to the understanding of the new rural cooperative medical system, design the data warehouse of the new medical insurance. And the data preprocessing such as the data extraction, cleaning and integration is completed. The medical insurance fraud identification of the new rural cooperative medical system is realized. And compare the effect of the different algorithms.