Research On UBI Pricing Model Based On Naturalistic Driving Data

The UBI(Usage-Based insurance)vehicle insurance is a hot spot in the research field of vehicle insurance at this stage.The mass naturalistic driving data can provide the driver's running data for the vehicle under the least disturbed condition for the UBI vehicle insurance research.After a reasonable analysis of the risk of vehicle operation data,the UBI vehicle insurance rate can be determined according to the risk.Because the price of vehicle insurance is determined by the behavior of the user,it is helpful to standardize the driver's driving behavior and improve the safety of road traffic by driving behavior to determine the cost of vehicle insurance.Therefore,this study seeks to build UBI pricing model reasonably based on naturalistic driving data,and to improve drivers'driving behavior by premium reward or punishment,and thus improve the level of traffic safety.First of all,on the basis of summarizing and analyzing the related achievements of traffic large data mining preprocessing,the application of naturalistic driving data,the theory and method of driving behavior research,the method and model of traffic risk assessment,the structure and characteristics of OBD data are introduced and analyzed,and the identification and optimization of data quality are also designed.This study uses SQL server and Python to complete data preprocessing,which facilitates the subsequent data analysis process to read and invoke the data.Secondly,the risk index in the claim frequency prediction model is determined by the method of theoretical analysis,and the mining algorithm of risk factors is designed.Python programming is used to complete the data mining and synthesize the vehicle reports containing each risk index.By using principal component analysis and factor analysis,the number of risk factors is reduced,and then this study finds that 5 groups of risk factors exist collinearity through multiple collinearity test,which provides the basis for the selection of variables for model construction.Thirdly,the Logistic model,Poisson regression model,negative binomial regression model,zero inflation Poisson regression model and zero inflation negative binomial regression model are used to construct the claim frequency prediction model,and the evaluation results show that the best fitting model is the Poisson regression model and the prediction accuracy of the model is 98.9398%.Finally,3444 vehicles are selected in the original database for example analysis,and the claim frequency prediction model based on Poisson regression is applied to the example of UBI self-determination premium coefficient adjustment,and then the rate coefficient of vehicle insurance is determined and the UBI vehicle insurance pricing process is completed.And we show the results of claims of the example.Then the premium of UBI vehicle insurance and the premium of traditional vehicle insurance are compared,and the rate change based on the UBI pricing is demonstrated.It is proved that the mechanism of reward and punishment for the drivers can be formed through the pricing of UBI.thus effectively improving the road traffic safety level.