Research On Regular Travel Behavior And Travel Modes Based On Urban Private Cars Trajectory

With the sustained and rapid development of China's urban economy,the contradiction between the rapid growth of vehicle ownership and limited urban road resources is increasing.Among them,private cars have a higher proportion.According to statistics,as of the end of 2017,private cars accounted for about 80.3%.On the one hand,the high ownership of private cars aggravates urban problems,such as energy consumption,traffic congestion and environmental pollution.On the other hand,with the development of positioning technology,information processing and data mining technologies,private cars provide large-scale trajectory data for Urban Computing,Intelligent Transportation Systems and other applications.Accurately collecting and acquiring private car trajectory data,analyzing and mining its unique movement patterns and trajectory characteristics are of great significance for studying the characteristics of residents' travel and the evolution of traffic flow.Based on the trajectory of urban private car,this paper studies the regular travel behavior and travel mode of urban private cars.The main work of this paper is as follows:There are data redundancy,drifting and other problems in the acquired private car trajectory.In order to eliminate the adverse impact of data quality problem on subsequent technical links,this paper carries out data cleaning on the original trajectory of private car.Frist,in order to reduce the data storage space,this paper removes redundant and invalid GPS(Global Positioning System)data.Then,in order to improve the performance of trajectory data analysis and mining,this paper uses threshold test method to detect and eliminate the drifting GPS data.Based on the historical trajectory data of urban private cars,this paper proposed a method to mine the regular travel behaviors of private cars.Since the similarity of trajectory can reflect the regularity of trajectory,first,this paper proposed IERP(Improved Edit Distance with Real Penalty)to measure the distance between trajectories,and then obtained the trajectory similarity matrix of private car.Then it utilize Kernel Principal Component Analysis(KPCA)to reduce feature dimension for enhancing model's performance and improving calculation speed.Finally,based on the transfer learning method,this paper mines the regular travel behaviors of private cars for solving the problem that a few labeled sample data is difficult to learn.This paper conducts regular travel behaviors mining experiments based on real urban environments' private car trajectory data.Based on the regular travel trajectory of private cars on work day,this paper proposed a method to mine the popular travel modes of private cars,which is the time allocation mode of private cars to different places.In order to facilitate the representation of the travel time feature,first,the travel stay information of the private car is extracted by using I-DBSCAN(Improved Density-Based Spatial Clustering of Applications with Noise).Then,according to the stay information,bar time blocks is used to effectively represent the travel time feature.Finally,this paper introduces ED into DBSCAN to measure the sample distance,and use ED-DBSCAN algorithm to mine the popular travel modes of the regular travel private cars.Based on the regular travel behavior of Shenzhen private cars,this paper carries out data mining experiments on its travel mode.