Application Of Zero-adjusted Regression Models And Credibility Theory In Calculating Auto Insurance Premium

Nowadays,motor vehicle insurance has become one of biggest property insurance business in China.In the area of the motor vehicle insurance,the most critical parts are premium rate determination and premium calculation.Especially in the era of today’s big data,the pricing methods that change from actuarial methods to large data modeling method have become an international trend.So how to make full use of large data information to find a more accurate pricing than traditional actuarial risk pricing,and how to develop more personalized exclusive products according to the characteristics of custom are key problems in the motor vehicle insurance business.Claim amount is one of the key indices that are considered in auto insurance premium pricing.Generally zero-inflated and the phenomenon of thick tail exist in the claim data.Compared to the traditional linear regression,the quantile regression that is more relaxed in assumptions about the data distributions has better robustness and can provide more information,so some scholars have introduced zero adjusted quantile regression model to analyze the data of motor vehicle claims.er segmentation,are the key problems in the motor vehicle insurance business.Credibility model is an empirical estimation model.With the development of Actuarial Science,the classical credibility theory has been constantly extended.Quantile credibility model is one of its extension forms.In this paper,zero adjustment model and credibility model are used to evaluate the premium of vehicle insurance.Based on logarithm linear quantile regression,firstly,we discuss two premium pricing methods in the context of the quantile premium principle: quantile premium and quantile credibility premium,and pure premium calculation based on machine learning regression algorithms is also exhibited.And pure premium based on other machine learning regression algorithms is exhibited too.We hope that the methods presented in this paper can provide some reference to the motor the vehicle insurance business.In addition,we also use quantile regression to fit and forecast the data of social insurance expenditure in China.The data processing in this paper is based on open source software-R.