Research On Automobile Insurance Fraud Identification Basedon Data Mining Method

In recent years,the insurance industry has been developing vigorously.However,the insurance fraud has gradually increased.The widespread existence of fraud causes the current situation of "high premium,high pay,low income" in insurance market.This has hindered the sustainable development of the whole business.Therefore,identifying insurance fraud,preventing and controlling risks and reducing losses not only have great guiding significance for the management of insurance companies,but also have great practical significance for promoting economic development and maintaining social stability.This paper firstly analyzes the current research situation of insurance fraud,systematically summarizes the concept,classification,performance and harm of automobile insurance fraud.Secondly,in view of the limitations of the existing research,a automobile insurance fraud identification system based on data mining technology is designed.Then,based on the theory of automobile insurance fraud identification,the model of automobile insurance fraud identification is established on Extreme Learning Machine and Random Forest.On the one hand,the variance estimation for traditional Extreme Learning Machine is not robust,so following the robust estimation theory,we propose the Extreme Learning Machine based on robust estimation.An iterative weighted least square method based on M estimation is used to robust estimate the output weight.At the same time,the parameters of Extreme Learning Machine are optimized by Bee Colony Algorithm,which can further improve the classification performance and generalization ability of Extreme Learning Machine,and simplify the network structure of the model.On the other hand,based on Balanced Random Forest,weighted comprehensive ranking method for evaluating the importance of variables is proposed.It integrates the feature importance scores of Random Forest and statistical t test as heuristic information.The Ant Colony Algorithm is used for intelligent search to improve the precision of the combined classifier.Finally,the index system of automobile insurance fraud is constructed based on the data of automobile insurance claims.The automobile insurance fraud identification model is established on Robust Extreme Learning Machine optimized by Bee Colony algorithm and Random Forest optimized by Ant Colony algorithm to mine the fraud law,identify the fraud and put forward some suggestions to help the management of insurance industry.