| With the rapid increase in the number of motor vehicles in China,auto insurance premiums have increased from more than 40 billion yuan in 2000 to an estimated 786 billion yuan in 2018,an increase of 18 times in 19 years,accounting for more than 75% of property insurance premium income.Affected by the deceleration of automobile consumption in 2018,the increase in auto insurance premiums in 2018 was less than 10% for the first time,the lowest point of auto insurance growth in 2000.At the same time,the insurance premium for commercial insurance vehicle insurance in the country was 614.9 billion yuan in 2018,an increase of only 2.6%.At the beginning of 2019,the National Development and Reform Commission issued policies to promote the consumption of new energy vehicles,rural commercial vehicles and used vehicles,which will certainly promote the growth of automobile sales,thus driving the development of automobile insurance.As the main insurance source of property insurance companies,auto insurance is an important source of premium income,but it has always been in a state of “high payout,low return”.In 2018,the comprehensive cost ratio of the auto insurance industry was 99.86%,the comprehensive expense ratio was 43.16%,and the comprehensive loss ratio was 56.7%.Due to the huge proportion of auto insurance business,the high cost of property insurance companies due to auto insurance fraud has made the comprehensive insurance premium rate remain high.According to statistics,auto insurance fraud accounts for nearly 80% of insurance fraud cases.Auto insurance fraud not only hinders the normal development of the auto insurance market,but also disrupts the public order of the society.It must guard against and crack down on auto insurance fraud.With the increase of anti-fraud,the forms of fraud are increasingly diversified and concealed.With traditional identification methods,it is difficult to effectively prevent auto fraud.Leveraging big data technology to realize auto insurance fraud identification and prevention is one of the important channels to break through the car insurance anti-fraud bottleneck.China's property insurance companies and regulatory agencies attach great importance to the prevention of auto insurance fraud.They have established and used auto insurance information sharing platforms to set up anti-fraud systems.However,due to the lack of overall planning,industry standards and data security regulations still need to be improved.Big data and auto insurance fraud The recognition model combination is still in the exploratory stage.Therefore,based on the background of big data,the use of supervised and unsupervised machine learning methods to study the risk and prevention of motor vehicle insurance fraud has certain innovative and important theoretical and practical significance.This paper adopts the combination of normative analysis and empirical analysis.Firstly,it starts with the related theories of motor vehicle insurance and insurance fraud,defines the concept of auto insurance fraud,and compares it with moral hazard.It analyzes the risk of auto insurance fraud according to the characteristics of motor vehicle insurance.Characteristics;brief introduction to the meaning,characteristics and key technologies of big data;secondly,analyze the types,performance,causes,hazards and prevention of auto insurance fraud risks according to the fraud subject,and briefly summarize the traditional and new measures to prevent auto insurance fraud risks Based on the big data characteristics,the in-depth analysis of the fusion point of big data and the risk of car insurance fraud prevention;then compare the advantages and disadvantages of the traditional common models and emerging models to identify auto insurance fraud,and select the supervised learning kNN algorithm in machine learning based on the big data background.And the unsupervised learning K-Means algorithm for the empirical research of vehicle risk fraud identification and the empirical results are compared;then through the comparative analysis of the US experience of applying big data to prevent auto insurance fraud and Chinese practice,it is concluded that it is in line with China's national conditions;most After summarizing and summarizing the research conclusions,this paper puts forward some countermeasures and suggestions for the application of big data in the field of automobile insurance anti-fraud in China,and makes research prospects.Based on the above ideas,the author divides the paper into six parts:The first part introduces the background,research purpose and significance of the thesis.It sorts out the research results of fraud risk and auto insurance fraud identification at home and abroad,and finds that scholars generally from the perspectives of information asymmetry,insurance contract,game theory and law.Research is carried out,and China's research on insurance fraud mostly focuses on the characteristics of insurance fraud and fraud prevention.There are few empirical studies on vehicle risk fraud identification and prevention.Foreign scholars have rich empirical research on insurance fraud,especially auto insurance.Medical insurance fraud identification,mostly using Logit,Probit,SVM,decision tree,Na?ve Bayes and other models,has recently begun to use more complex methods such as neural networks,gradient lifting decision trees.Therefore,this paper is based on the big data background to use the US AIB simulation data set to use the supervised and unsupervised machine learning methods to study auto insurance anti-fraud is of exploratory significance.The second part gives an overview of the concept,characteristics,market subjects and their particularities of motor vehicle insurance.It also sorts out the information asymmetry theory,behavioral economics,institutional economics and economic ethics related to insurance fraud.The theoretical basis;the concept of auto insurance fraud is defined by comparing with moral hazard,and its characteristics are analyzed;the meaning,characteristics and key technologies of big data are briefly introduced,which will pave the way for the introduction of big data into auto insurance fraud risk prevention.The third part summarizes the main manifestations and types of motor vehicle insurance fraud from the perspective of auto insurance fraud,and analyzes the causes,harms and precautions of auto insurance fraud risks,and briefly summarizes the traditional and new measures to prevent auto insurance fraud risks.Data characteristics provide an in-depth analysis of the convergence of big data and the risk of fraud prevention.Since the nature of the auto insurance data itself is compatible with the big data characteristics and the use of big data technology can improve the accuracy and efficiency of auto insurance fraud,it is feasible to combine the big data to prevent the risk of auto insurance fraud,but the whole process from data collection to processing exists in practical applications.Difficulties,and data security and privacy protection issues cannot be ignored.The fourth part compares the advantages and disadvantages of traditional common models and emerging models of auto insurance fraud.Based on the big data background,the supervised learning kNN algorithm in machine learning and the unsupervised learning K-Means algorithm are used to empirically identify auto insurance fraud.Comparing the empirical results,it is found that the accuracy of fraud detection of the two algorithms is comparable.However,in the field of auto insurance and anti-fraud based on big data background,unsupervised machine learning can achieve more accurate and extensive fraud detection,and has more development potential.The best anti-fraud method is to combine the anti-fraud expert rules with the machine learning model.The fifth part introduces the American experience and Chinese practice of applying big data to prevent car insurance fraud risk.Through comparative analysis,it is found that domestic anti-fraud system service providers are not inferior to developed countries such as the United States in artificial intelligence technology,but they are built on data platforms.There is still a distance between data sharing and data security protection.In addition,the experience of the United States in the construction of professional anti-insurance fraud institutions,anti-insurance fraud law formulation and anti-insurance fraud propaganda is worth learning from.The sixth part summarizes the research results of this paper.On this basis,it proposes countermeasures and prospects for future research.There are two innovations in this article.First,the topic selection and research perspective are innovative.The author attempts to explore the risk and prevention of motor vehicle insurance fraud based on the background of big data.It not only analyzes and compares the advantages and disadvantages of traditional car insurance fraud risk identification model and emerging model,but also deepens the integration point of big data and car insurance fraud risk.This paper discusses and analyzes the case of applying big data anti-auto insurance fraud in China and the United States to draw lessons from China's national conditions.It has certain theoretical and practical significance for the application and development of big data in the field of automobile insurance anti-fraud.The second is to use the cutting-edge machine learning method for empirical research.Through self-learning Python and machine learning basics,the author tries to use the machine learning method under the big data framework to conduct empirical research on vehicle risk fraud risk identification.Select supervised learning kNN classification algorithm and unsupervised learning K-Means clustering algorithm for the United States respectively.The AIB simulation data set is used to identify auto insurance fraud,and the results of the analysis are compared.It is concluded that the use of unsupervised learning methods in the context of big data is more advantageous. |
PDF Full Text Request