【作者】 高岩；

【导师】 刘志硕；

【作者基本信息】 北京交通大学 ， 物流工程（专业学位）， 2017， 硕士

【摘要】 我国机动车保险行业规模巨大,然而盈利状况却一直不甚理想,其根本原因在于我国车险定价长期以来都遵从从车因子,忽略了驾驶员的驾驶行为差异。UBI(U sage based Insurance)作为一种欧美发达国家新兴的车辆保险模式,能够根据驾驶员的驾驶行为安全程度个性化地确定保费,从而约束驾驶员不安全驾驶行为,降低事故率,减少理赔。在我国新的费改政策环境下,研究基于我国新费改政策的UB I费率厘定理论与方法具有重要的理论价值与实际意义。因此,本文在UBI国内外研究与应用现状的基础上,提出了基于我国新费改政策的UBI费率厘定模型。首先使用分类算法建立驾驶行为与事故风险的联系,然后根据分类结果确定UBI核保系数,再将UBI核保系数引入新费率政策下的保费计算公式,形成适合我国国情的基于驾驶行为分类的UBI费率厘定模型。论文的主要内容如下:(1)建立基于驾驶行为分类的UBI费率厘定模型。首先介绍了车险费率厘定的理论基础,费率厘定因子和费率厘定方法。分析指出了现有UBI费率厘定理论的不足,即无法对驾驶行为进行客观评价。在此基础上,提出使用数据挖掘的分类技术代替依靠主观赋权法建立的驾驶行为评分体系对驾驶行为进行评估。并在新的费率体系下,建立了基于驾驶行为分类的UBI费率厘定模型,给出了费率厘定公式与厘定步骤。最后将模型应用于保费实例计算,计算过程中证明了驾驶行为分类模型对于事故风险的识别能力确实优于驾驶行为评分模型,可以使UBI费率厘定更加科学、合理。(2)根据车联网数据建立驾驶行为分类模型。以400名驾驶员的实际驾驶行为数据和出险数据作为实验数据,首先使用ExtraTrees算法进行特征选择,然后分别采用决策树、朴素贝叶斯、k-NN、神经网络、SVM分类器进行训练,再根据训练集和测试集上的准确率、测试集上的混淆矩阵和ROC曲线对分类器进行选择,得到分类性能最优的基于SVM分类器的驾驶行为分类模型,并用PSO算法与GA算法对模型参数加以优化。更多还原

【Abstract】 China has a huge auto insurance market,but the profit of the industry is not satisfactory.The main reason for this situation is that China’s auto insurance pricing only considered the car factor,ignoring the driver’s driving behavior differences.UBI(Usage Based Insurance)is an emerging auto insurance model of Eur opean and American countries,which determine the premium according to the driver’s driving behavior safety level to restrict the driver unsafe driving behavior,reduce the accident rate and claims.Under environment of the rating policy reform of our country,it is of great theoretical and practical significance to study the theory and method of UBI rating based on China’s new rating policy.Therefore,based on the present situation of UBI research and application,th is paper puts forward the UBI rating model based on driving behavior classification.Firstly,the classification algorithm is used to establish the relationship between driving behavior and accident risk,and then the UBI coefficient is determined according to the classification result.Next,the UBI coefficient is introduced into the premium calculation formula underwriting new rating policy.A UBI rating model based on driving behavior classification for our country is established.The main contents of the paper are as follows:(1)Established UBI rating model based on the driving behavior classification.Firstly,the paper introduces the theoretical basis of the auto insurance rating,consist of the rating factor and the rating method.The follow analysis points out the shortcomings of the existing UBI rating theory,that is,the objective evaluation of driving behavior can’t be carried out.So the idea of using the classificati on technology of data mining replace driving behavior scoring system which relying on the subjective weighting method to evaluate the driving behavior was pro posed.Then the UBI rating model based on the driving behavior classification u nder the new rating reform policy was established,and the premium calculation formula was determined.Finally,an example shows that the driving behavior classification model is superior to the driving behavior score model for the identification of accident risk,which can make UBI rate more scientific and reasonable.(2)Determined the driving behavior classification model according to the telematics data.The actual driving behavior data and the risk data of 400 drivers are used as experimental data.Firstly,the feature selection is performed using the ExtraTrees algorithm,and then trained the Decision tree,naive Bayes,k-NN,neural network,SVM classifier use selected features.According to the accuracy of the training set and the test set and the confusion matrix and the ROC on the test set,the driving behavior classification model based on SVM is determined.Then the model parameters was optimized by PSO algorithm and GA algorithm.更多还原