中文摘要：

在车险费率厘定中，广义线性模型的应用最为广泛，然而其假设散度参数为常值，这限制了模型在车险定价中的进一步推广。双重广义线性模型能够对反应变量的均值与散度参数同时建立广义线性模型，提高了模型运用的灵活性与适应性。本文将双重广义线性模型应用到车险费率厘定中，直接对纯保费建立模型，以一组实际的汽车保险损失数据为样本进行实证研究，并与常用分布假设下的广义线性模型进行了对比分析。结果表明，双重广义线性模型得到的费率结构更为合理，符合实际。

英文摘要：

Generalized linear models are widely used in auto insurance ratemaking. However, their assumption of dispersion parameter being constant prevents them from further adoption in the pricing of vehicle damage insurance. Double generalized linear models can be used to model the means and dispersion parameters at the same time, im- proving model flexibility and adaptability. This paper applied double generalized linear models to vehicle insurance ratemaking,whose pure premiums were modeled directly. An empirical study on a realdataset of automobile insur- ance loss was carried out, and the result was compared with that of traditional generalized linear models under as- sumptions of common distributions. The study showed that the rate structure obtained by our model was more reason- able and practical.