Research On The Model Of Auto Insurance Risk Customer Segmentation Based On K-PrototypesNN Method

In recent years,the domestic auto insurance market has made considerable progress,and is also an increasingly competitive market.In the case of auto insurance product homogenization and competitive,many insurance companies pursue the market scale unilaterally and ignore the growth of auto insurance benefits,which seriously affects the sustainable development of the auto insurance industry.The main obstacle of the auto insurance business at present is the unification rate standard of the auto insurance market,and the same rate for different customers,which directly cause the loss of insurance company.Therefore,the insurance company must change the mode of operation,take the road of refined management of auto insurance customers.With arrival of the era of big data,the per capita car ownership in the market continues to increase,and auto insurance consumers will increase,resulting in a large number of data in the auto insurance market.Auto insurance data has the characteristics of dynamic,many types and large amount.Therefore,mining from data to research the common characteristics of customer group,and using the massive data mining more behavior information has important practical significance to solve customer orientation,risk partition problem,solved the problem that current risk management method can not be accurate positioning of risk of the customer base,and have certain reference value for theoretical research and practice.In this paper,the main work of auto insurance risk customer groups as follows:(1)This paper aims at the problem that the domestic insurance companies pay too much attention on the premium and ignore the extensive management mode,studies the set of foreign auto insurance classification rate,analyzes the reasons of profitability of domestic insurance companies is not ideal and analyzes the problem of high auto insurance rates and imperfect rate setting system.(2)This paper systematically analyzes the necessity of the division of auto insurance risk customer groups at three aspects.For auto insurance payment rate is high,the insurance company classify customer risk level extensively and traditional risk management methods are hard to achieve the segmentation respectively.(3)This paper analyzes the risk factors of auto insurance,including the driver,vehicle and environment 21 factors,according to the characteristics of auto insurance customer data belong to the mixed type data,this paper proposes a K-Prototypes NN algorithm,and tested in the UCI data,the experimental results show that KPNN in accuracy and macro average recall 4 indicators are better than the K nearest neighbor algorithm.In this paper,a model of auto insurance customer division based on KPNN algorithm is proposed.(4)In the empirical part,through questionnaires,7153 questionnaires were distributed,and the effective datas were 6916.And the risk of auto vehicle insurance customers are divided into high,medium and low three categories.2/3 datas are used as training samples,and 1/3 datas are used as test samples.This paper classifies customer groups,the vehicle driver risk factors,customer groups risk factors,customer risk factors,the driver part of customer group and vehicle part of group factors affecting customer factors respectively,the prediction accuracy was above 70%.Which proves the feasibility of the data mining algorithm for the division of auto insurance risk customer groups,and then achieve the purpose of auto insurance risk rate segmentation.(5)In order to realize the differential pricing of auto insurance rates,this paper puts forward some suggestions for the sustainable development of the insurance market based on the data mining technology.To speed up the process of auto insurance rates market,to construct the risk of customer segmentation model,to establish an open and uniform auto insurance data sharing platform and improve the regulatory system,respectively.