Category Archives: Parametric Insurance

【Abstract】 The basic Principle on car insurance pricing is risk analysis. The conditious of car, the car drivers,and circumference of driving make a whole risk system of car insurance pricing. most of the traditional pricing method for car insurance are deterministic, such as single analysis, the minimum deviation method, Multiple linear regression model and so on. The deterministic method has the advantage lies in the process simpler, but can not reflect the random effect of risk.so people try to introduce the stochastic model to the car insurance pricing. The Generalized Linear Model(GLM) has been used widely in the these stochastic models, Generalized linear model has a greater progress in the theory, model structure and model of diagnosis, It is more close to real life and more wider field of use. Although the GLM to solve the limitations of traditional linear model, but GLM system components still showed a linear form of covariates, in many cases, sequence data, especially spatial data, covariate effects often show a strong nonlinear, which makes the application of GLM subject to certain restrictions. We introduce the Generalized Addictiv Model (GLM) to car insurance pricing as a starting point. GAM is an extension of GLM, It retains the basic framework of GLM, set part of the covariate effects on the linked as nonlinear functions, and applly non-parametric smoothing technique to the exponential family data.this article is divided into six parts. The first part introduces the background, significance and research methods. The seconde part describes the traditional car insurance pricing. The third and fourth parts describe the generalized additive model, which apply the nonparametric smoothing to estimates parameter, combined with bayes theory and MCMC methods. The fifth part construct the Gamma Addictiv Model and Poisson Addictiv Mode for the car insurance. The sixth part are the empirical part, which construct car insurance pricing model to a foreign insurance data.The empirical results that this method is more suitable for the non-linear covariate and will provide more reference space for car insurance pricing, despite the extensive application of GAM should be study furtherly.