面向车险定价场景的车型风险分级研究

Research on Vehicle Type Risk Rating for Auto Insurance Application Scenarios

本文以保险公司理赔数据及车型配置数据为研究对象,基于GLM模型和K-means分级方法对我国车型风险进行分级。首先,使用GLM模型对车参配置因子进行筛选,确定对理赔数据有显著解释性的因子,并得到理赔金额拟合值。其次,使用K-means聚类算法,对拟合结果进行等级划分,最终将车型风险划分为30个等级。结果表明,车型风险分级各等级分布直方图呈现伽马分布,符合车险定价分布特征。该研究内容及模型算法,将为我国汽车车型风险分级体系构建工作提供重要方法参考,并为车险定价创新提供支持。

In this paper,insurance claims data and vehicle configuration data as the study objects,based on GLM model and K-means classification method of Chinese vehicle risk rating.Firstly,the GLM model was used to screen the vehicle parameter configuration factors,determine the factors that have significant explanatory effect on the claims data,and obtain the fitting value of the claims amount.Secondly,the K-means clustering algorithm was used to grade the fitting results,and finally the vehicle risk was divided into 30 grades.The results showed that the histogram of each grade of vehicle type risk classification showed gamma distribution,which was consistent with the distribution characteristics of auto insurance pricing.The research content and model algorithm will provide an important reference for the construction of the risk classification system of automobile models in China,and provide support for the innovation of automobile insurance pricing.