Research On Auto Insurance Fraud Identification Based On Fuzzy C-means And Logistic Regression Model

With the vigorous development of the insurance industry,the problem of insurance fraud has attracted more and more attention.Auto insurance fraud has always been a serious category of all insurance fraud,which is very important to the development of the insurance industry.Therefore,auto insurance fraud identification technology has always been a hot issue studied by scholars at home and abroad.In view of the late start of the research on this problem in China,the technology of insurance fraud identification is not very mature.However,industry personnel attach great importance to this.In recent years,there have been many research results and extensive practical applications.As for the auto insurance fraud identification model,the traditional method is the logistic regression model,and the machine learning method is widely used in recent years.The logistic regression method has good explanatory ability.However,with the development of big data and the Internet,the defects of logistic regression model,such as too many factors,too high dimensions,increased dependence between variables,time-consuming modeling and reduced prediction ability,have become prominent in practical application.In this thesis,the fuzzy c-means algorithm is introduced to cluster multiple factors related to auto insurance fraud according to the explanatory variables related to accidents,people and other factors,so as to reduce the dimension,and then establish a logistic regression model for fraud identification.In this thesis,the auto insurance fraud identification method based on fuzzy c-means and logistic regression model combines the strong interpretability of traditional generalized linear model method and the effectiveness of modern machine learning method in dealing with high-dimensional complex data,which improves the efficiency of auto insurance fraud identification.The results of case analysis show that the method proposed in this thesis has certain advantages in auto insurance fraud identification.Compared with the existing methods,the innovation of this thesis is that it combines the traditional method of logistic regression model and the modern method of machine learning,which provides a new idea for the research of auto insurance fraud identification and improves the identification efficiency.The structure of this thesis is as follows:firstly,it introduces the background and research significance of auto insurance fraud,and explains that identifying auto insurance fraud has profound practical significance.Then,the concepts and principles of fuzzy c-means method and logistic regression model used in this thesis are briefly described.Next,a group of auto insurance fraud related data from kaggle website is selected for empirical research.The results show that the method in this thesis makes the model factors more significant and the prediction accuracy higher.Finally,it summarizes the full thesis and looks forward to the future work.