On the Application and Comparison of GLM and GAM in Modeling Claim Frequency of Auto Insurance
Duan Bai-ge · Xiandai caijing · 2012
The rate reform of auto insurance is a hot topic which draws more concern by insurance industry in recent years.The pricing and rate reform of auto insurance is based on risk analysis.Meanwhile,vehicles,drivers,driving environment and other factors constitute the risk system on which auto insurance pricing depends.In this paper,we introduce both GLM-logistic regression model and GAM-logistic regression models,and incorporate semi-parametric smoothing method into the analysis of impact factors in modeling claim frequency of auto insurance.Using a sample of overseas auto insurance data,we propose to construct the models in terms of impact factors on claim frequency of auto insurance,furthermore it provides a comparative study in terms of the two models.We make an empirical analysis on R software and predict occurrence probability of claims.The result shows that the GAM-logistic regression model based on semi-parametric method has more advantages than GLM logistic model,and also has a better prediction effect.