Research Of Insurance Fraud Monitoring Method Based On Data Mining Technology

With the development of the insurance industry, the diversification of insurancecategories and the increasing of the insurance charge, the insurance fraud is increasingyear by year. Vehicle insurance, as one of the property insurance, is developed widely bythe appearance of cars. Vehicle insurance fraud in recent years become more professionaland complicated. And the amount involved is increasing year by year. Because vehicleinsurance frauds, hidden and diversified, bring about the un-measurably negative impactto people. Therefore, this paper studies the methods of insurance fraud, the main contentsare as follows:Firstly, based on auto insurance data from China United Property Insurance Company,I selected eight properties, and the gave a systemetical analysis at the beginning. And thenwith the use of Logit measurement model、decision-making tree、Naive bayes andBayesian network model, a more comprehensive analysis to the data is given.Secondly, This paper studies the Complementarity of Integration algorithm on thedifferent characterized description and the different classifiers. Performance. With the useof the Integration algorithm of Bagging、AdaBoost、Majority Voting and stacking,monitoring models are established. Through comparing them, and then got one higerclassification accuracy model. This model can identify the insurance fraud effectively onvehicle insurance and improve the monitoring capabilities of insurance business.Finally，through a large number of comparative tests, I proposed a mining-methodsuit to insurance fraud and obtain good results. Based on this, policy recommendationsare given to prevent insurance fraud. This paper enriches the theory of insurance fraudmonitoring and provides effective analysis tools and decision support for the insurancecompany.