中文摘要：

广义线性模型和广义可加模型作为经典线性模型的扩展,近年来在非寿险精算中得到了广泛的应用.本文在对2种模型进行简介的基础上,将驾驶员的性别、车型等8个变量作为费率因子,分别建立了车险索赔发生概率估计的广义线性模型和广义可加模型,并选取瑞典瓦萨（Wasa）保险公司的车险数据对2种模型的估计效果进行比较分析.结果表明,对于离散型费率因子占绝大多数的车险数据,广义可加模型并不具有明显的优势.因此,在车险费率厘定实务中,若离散型费率因子较多,应选择结构相对简单的广义线性模型.

英文摘要：

As extensions of classical linear model, Generalized linear models and Generalized additive models recently have been widely used in non-life actuarial science. In this paper, by using eight variables including gender and vehicle type as the rating factors, the probability of claim is modeled applying Generalized linear models and Generalized additive models respectively. Furthermore, the estimation effects between the two models are compared by applying the data of Wasa insurance company of Swedish. It is shown that Generalized additive models does not has clear advantage in fitting the data of automobile insurance because of the existence of more discrete covariables. Therefore, Generalized linear models should be adopt in insurance practice when there are more discrete risk factors.