摘要
随着车辆网技术的不断成熟,车联网数据的应用价值日渐凸显。车联网大数据中包含着丰富的驾驶行为信息,这些信息对于改进传统的汽车保险定价模型具有重要的应用价值。如何从车联网大数据中提取出具有实际应用价值的信息,尚需进行大量细致的研究工作。本文基于车联网记录的速度-加速度数据,应用核密度估计和主成分分析,提取了一个驾驶行为因子,并在泊松分布假设下建立了索赔频率的广义可加模型。实证研究结果表明,本文提取的驾驶行为因子对被保险车辆的索赔频率具有十分显著的非线性影响,为汽车保险定价提供了一个新的费率因子,有助于进一步提高汽车保险定价结果的准确性和合理性。
With the development of vehicle telematics, the value of telematics data becomes apparent in insurance. Telematics car driving data contains detailed driving behavior information ,which may be used to improve the traditional rate making models in ear insurance. How to extract useful information from telematics data is an on-going research topic and it requires sophisticated statistical methods. This paper analyzed the speed-acceleration data from the telematics car driving data. Kernel density estimation and principal components analysis were applied to extract a driving style index from the speed-acceleration data, and a Poisson generalized additive model for claims frequen- cies was established. The empirical analysis showed that there was a quite significant non-linear relationship between the extracted driving style index and the claims frequencies. This driving style index can be used as a new rating factor, which will improve the predictive power of the rate making model and lead to a more accurate and reasonable rate.
出处
《保险研究》
CSSCI
北大核心
2018年第1期90-100,共11页
Insurance Studies
基金
教育部人文社会科学重点研究基地重大项目"基于大数据的精算统计模型与风险管理问题研究"(16JJD910001)
国家社科基金重大项目"巨灾保险的精算统计模型及其应用研究"(16ZDA052)
中国人民大学2017年度"中央高校建设世界一流大学(学科)和特色发展引导专项资金"支持
关键词
车联网
大数据
汽车保险
费率因子
驾驶行为
索赔频率
telematics
big data
car insurance
rating factor
driving style
claims frequencies