考虑到离群值对未决赔款准备金评估的影响,在双广义线性模型中引入离群值,建立稳健的双广义线性模型,对离群值进行识别和调整,使得模型参数在处理含离群值数据时,可以得到更加稳健的评估结果。针对含有离群值的赔付次数和案均赔款数据,分别建立了稳健的双广义线性模型和传统广义线性模型,得到相应的未决赔款准备金评估值。最后对比分析了2个模型对离群值的敏感性。结果显示:稳健的双广义线性模型可以对离群值进行有效的识别和调整,最终未决赔款准备金评估结果也比传统模型更加平稳。
Considering the impact of outliers on the outstanding claims reserve evaluation,the outliers was introduced into Double Generalized Linear model. The robust Double Generalized Linear model thus was established to identify and adjust outliers so that the evaluation results can be more robust when the outliers are contained in the model. According to the number of paid and payments per claims within the outliers,the robust Double Generalized Linear model and the traditional model were constructed and the corresponding outstanding claims reserve evaluation values were obtained. The results showed that the robust Double Generalized Linear model can effectively identify and adjust outliers. The final evaluation result of the Robust Double Generalized Linear model is more stable than that of the traditional model.