Generalized linear models assume that the influence of the independent variables on the response variable function is linear as a mainstream approach to the research of car insurance pricing. But in the actual situation, a number of factors that affect the frequency and severity of claims are not only a linear form. Simply using linear estimation will cause some variables insignificant, thus it is necessary to introduce generalized additive model. Based on a real dataset of automobile insurance loss, this thesis discussed car insurance pricing with different factors by using GAM-Tweedie model. It showed that the generalized additive GAM-Tweedie model can better explain the impacts on claims than the linear model.