Analysis And Research On Group Driving Behavior

With the number of urban operation vehicles increases,the Group's management of operating vehicles concerns public transport safety and the business benefits.Group car embedded devices generated a large amount of data.It's significant that analyzing and mining effective information from large volumes of data to ensure public traffic safety,accurate classification management,training of drivers and increase business efficiency.The paper first describes the application of data mining in driving behavior research at present both at home and abroad and the vehicle accident prevention system are introduced.Some pretreatment work of analyzing the original driving behavior data generated by the automobile accident prevention system are conducted including business analytics,the data cleaning,the feature generation and so on.In order to find out the similarity of driving behavior among different drivers.First,K-means algorithm and Gaussian Mixture Model algorithm(GMM)are used to analyze the driving behavior alarm information of the drivers in the city traffic in Beijing.This system products six kind of warning including lane departure,vehicle distance detecting,speed limit warning,collision in city,vehicle frontal impact and acceleration over threshold.According to the statistical alarm frequency of drivers'actual driving time and idling time to clustering analysis the similarity of different drivers in each dimension.Second,the similarity measure is made for the speed time series of the driver's continuous driving time and the series of the number of the alarm type per ten minutes by the Dynamic Time Warping(DTW)algorithm.Symbolic Aggregation Approximation(SAX)algorithm is used to symbolize time series,and combined with the suffix array on the alarm series and the continuous driving time of the speed time series for periodic pattern mining to analyze the similarity of the periodic pattern.Based on it,developing a car driving behavior analysis system.The group's chauffeur can be clustered and classified according to the driving behavior,which provides the basis for the group's targeted management and training drivers.