Cellular Data Based Travel Characteristics Analysis Method For Rail Transit Passenger

Urban rail,as a large capacity,long distance,safe and efficient travel mode,shares a great deal of travel flow in city and plays an important role in congestion releasing.As the increasing passenger flow,the transit is facing with the problems like optimization and improve.The management,planning and controls of transportation from the whole traffic system,by study on transit passengers' travel characteristics both in the rail system and outside,becomes a useful approach.As a kind of big data,cell-data has many advantages in the study of individual travel characteristics and group travel characteristics,and has shown potential application in the field of transportation.Based on the cell-data,both the inside feature and outside feature of transit passengers is analyzed in this thesis.Firstly,through literature research,the basic theory of cell-data generation and data structure are briefly introduced.And there is a brief introduction about the research object and a detailed description about the data that will be used in the research.Based on past experience,the noise in cell-data is analyzed and a method of ping-pong identification based on data sequence is proposed,which improves the method of three data recognition and can adat to different forms of ping-pong switiching in theory.Secondly,based on the analysis of rail travel,the rail-travel time is divided in to four parts,namely arrival time,traveling time,transfer time and departure time.Then a data relationship based recognition algorithm for key trajectory points such as rail travel enter-station,transfer and exit-station is presented.For the sensitivity of the previous algorithm in the travel time fluctuation,the dynamic time threshold analysis and calculation method is proposed,which can adapt to the travel time fluctuation of different travel distance.In order to handle the exception in real cell-data,like data break,data mergence,etc,the study also makes a corresponding analysis and processing.Thirdly,through the outside-trip analysis,the calculation about an additional residence time is introduced to improve the calculation of cell-data's residence time which is divided into two parts:the actual dwell time and the supplementary dwell time.Combining the trip recognition in the rail system,a dynamic distance threshold based trip start and end identification method is proposed to adapt to the difference of base station spacing in different regions.And an exploratory study,using the unsupervised method,is carried out to infer the transportation mode share outside the rail.Finally,the whole identifying process is described with an example of single user data.Combining the AFC data,the identification method of rail travel characteristics is evaluated.The result shows the proposed method is consistent with the AFC data.Additionally,the temporal and spatial distribution of cell's origin and destination is briefly analyzed.The paper summarizes the shortcoming and puts forward prospects for future research.