Research On The Travel Mode Recognition Method Based On Mobile Phone Signaling

With the acceleration of urbanization and motorization,the travel characteristics and rules are constantly changing.Analysis of such change provides a reference and guidance for transportation planning,urban management and travel navigation.Furthermore,the development of mobile positioning and wireless communication has made the research of using GPS signals,mobile phone signaling and other data to obtain wide-area travel information a hotspot.The mobile phone signaling data is generated in the mobile communication system and used to record the communications between the mobile stations and the base stations,implying the location information of users.Compared with GPS,it has a wider coverage and a larger subscriber amount,and it is easier to reflect the travel characteristics from a more macroscopic perspective.Therefore,this thesis uses signaling data to study the mining of travel modes.The current methods of using the signaling data for travel analysis cannot meet the accuracy requirements of the application scenario,because there are insufficient aspects in the processes of data preprocessing,trajectory segmentation and travel feature calculation.It is necessary to study a trajectory generation method and travel mode recognition method for the mobile phone signaling data.This thesis proposes a method on travel modes recognition based on mobile phone signaling data.The method is divided into two stages: trajectory generation and travel mode recognition.In the generation stage,firstly the time and space threshold filtering method are designed to identify and deal with the abnormals such as “ping-pong switching” and “data drift” caused by uneven distribution of base stations and terrain differences.Then the key points indicating “parking” and “transferring” are identified through feature analysis and used to segment the trajectory,so that one track can be divided into multiple sub-tracks reflecting different travel modes.In the recognition stage,by introducing the road network information,constraining the distance of the signaling points,and improving the calculation method of the feature values,the problem of low accuracy in position caused by the cellular positioning is solved.Then the travel feature data set is generated by calculating the selected feature values and used to be analyzed by an unsupervised machine learning method.Finally,the travel modes represented by various clusters are recognized by the analysis of the original signaling trajectories' feature distribution.In this thesis,the proposed method is experimentally verified by using an actual user mobile phone signaling data set.The experimental results show that the method can effectively recognize the users' travel modes based on the mobile phone signaling data.In the trajectory generation stage,the abnormal situations and key points in the original data can be accurately identified.The processed tracks are smoother and clearer without affecting the users' travel characteristics,and the segmented sub-tracks are reasonable.In the travel mode recognition stage,the method can effectively improve the accuracy of the travel feature values by conducting the road network constraints,and the recognized travel modes are reasonable and effective.