加入中国保险汽车安全指数指标的车险索赔次数研究

在车险定价模型已改进、汽车碰撞测试结果已发布的背景下，研究不同模型以及纳入模型中解释变量的组合方式对索赔次数预测效果的影响。由于广义线性模型（Generalized Linear Models,GLM）存在局限性，仅包括线性预测部份，对连续型变量解释性较差，所以采用了广义线性可加模型（Generalized Linear Additive Models,GAM）对车险数据进行拟合，同时研究中国保险汽车安全指数（China Insurance Automotive Safety Index,C-IASI）指标的3种不同风险因子组合方式对模型效果的影响。结果表明，广义线性可加模型对车险索赔次数的拟合效果优于广义线性模型；维修经济性得分与耐撞性得分作为自变量时的拟合效果优于其他组合。

Research on the Number of Insurance Claims Considering China Insurance Automotive Safety Index

Based on the pricing models and the vehicle crashworthiness ratings, the paper studies the impact of different models and the combinations of explanatory variables in the model on the prediction of the number of insurance claims. The generalized linear models (GLM) only produce the linear prediction and poorly interpret continuous variables. Therefore this paper used a generalized linear additive model (GLM) to fit the auto insurance data, and studied the impact of three combinations of C-IASI risk factors. The results show that the fitting effect of GLAM on the number of insurance claims is better than that of the GLM. When reparability score and damageability score are used as independent variables, the fitting effect of which is better than that of other combinations.