The Study On The Rating Of Automobile Insurance With Human Factors

The reform of Auto insurance rates was carried out in 2003,many experts believe that the auto insurance's rating must consider adding the variable factors about drivers,such as the age, sex, marital status, and violation records of the drivers. But in fact, The majority of property insurance companies in China have not record these factors in detail. The automobile insurance rating model lack of those factors is not accurate.The paper focus for automobile insurance in China,which factors about drivers will significant impact the claims.Through single-factor analysis on Sample data we can concluded that:The younger drivers have the higher relative risk, the group of 18-25 age is the highest.The risk of male drivers is higher than women's.The risk of Unmarried drivers is higher than married drivers.public auto's risk is higher,and enterprises's is higher than Institutions'.The renewal divers'risk is lower than the divers of unwilling renewal.The drivers whose driver age is less than one year have the greatest risk.The driver age is between 3 and 6 years have relatively stable risk , between 3and 4 years have the lowest risk. After that,driving risks have the trend of increasing With the growth of drive age.Reviewing the automobile insurance's rating model, we found that compared with other models, the generalized linear model mainly be used the regression analysis of the Non-Normal data. In the insurance industry, this model is not only retain the traditional advantages of the linear regression, but also make loss of distribution's modeling easier.The last part of the this paper will divide all drivers into several groups according different rating factors,and determine those groups'premium.Rating automobile insurance must consider loss frequency,loss severity , additional costs and other financial factors.Take a case for example,the paper respectively discuss about the generalized linear model modeling of loss frequency and loss severity,from the statistical analysis to model selection and diagnosis.