基于累积索赔金额的最优奖惩系统

发布时间：2019-03-06 18:18

【摘要】：商业汽车保险的费率可以分解为先验费率和后验费率。先验费率是基于被保险车辆的先验风险特征信息(如驾驶人的性别和年龄、车辆使用性质)应用广义线性模型厘定的费率,而后验费率是基于被保险车辆的索赔经验(如索赔次数)应用奖惩系统对先验费率进行调整而得到的费率。厘定先验费率常用的广义线性模型包括索赔频率模型、索赔强度模型和累积索赔金额模型。实际应用中的奖惩系统通常基于被保险车辆的经验索赔次数对先验费率进行调整,没有考虑索赔金额的影响,也没有考虑先验信息的影响,有可能造成重复性的奖励或惩罚。累积索赔金额是被保险车辆在保险期间的索赔金额之和,既包含索赔次数信息,也包含索赔金额信息,可以更加准确地揭示被保险车辆的索赔经验信息。本文应用被保险车辆多年期的纵向累积索赔金额数据建立了一种新的奖惩系统,应用零调整逆高斯回归模型厘定先验费率,并在线性约束下用极大似然法同时估计先验费率因子和奖惩系数,避免了传统奖惩系统对被保险人可能造成的重复性奖励或惩罚,有效改进了后验费率的厘定结果。
[Abstract]:The commercial car insurance rate can be broken down into the prior rate and the posterior rate. The premium rate is based on the prior risk characteristics of the insured vehicle (such as the sex and age of the driver, the nature of the vehicle use) and applies the generalized linear model to determine the rate. The posterior rate is based on the insured vehicle's claim experience (such as claim times) and the premium and punishment system is applied to adjust the prior rate. The generalized linear models commonly used to determine the prior rate include the claim frequency model, the claim intensity model and the cumulative claim amount model. In practice, the reward and punishment system usually adjusts the priori rate based on the number of the insured vehicle's experience claim, without considering the influence of the claim amount or the priori information, which may result in repetitive reward or punishment. The cumulative claim amount is the sum of the insured vehicle's claim amount during the insured period, including both the number of claims and the claim amount information, which can more accurately reveal the insured vehicle's claim experience information. In this paper, a new reward and punishment system is established by using the multi-year longitudinal cumulative claim amount data of insured vehicles. The zero-adjustment inverse Gaussian regression model is used to determine the prior rate. The maximum likelihood method is used to estimate the priori rate factor and the reward and punishment coefficient simultaneously under the linear constraint, which avoids the repetitive reward or punishment caused by the traditional reward and punishment system to the insured, and improves the result of determining the posterior rate effectively.
【作者单位】：中国人民大学应用统计科学研究中心;中国人民大学统计学院;
【基金】：国家社会科学基金重大项目“巨灾保险的精算统计模型及其应用研究”(16ZDA052) 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)的资助
【分类号】：F840.63

【相似文献】