Research On Urban Public Transit OD Demand Estimation Method Based On Multi-source Data Fusion

Mining public transit OD demand based on multi-source data fusion in urban public transportation system if of great significance for public transit planning and operation management.Under the “unified fare” charge system,only the passengers' swiping time,bus plate number and some other charge information can be obtained from IC card data,and the passenger's board site cannot be directly obtained,which some relevant inference acquisition is required.Moreover,the accuracy and recognition rate provided by the existing research are not high enough for passengers' depart site and depart time,which leads to the difficulty for analyzing the spatial-temporal distribution characteristics of passengers flow and obtaining public transit OD demand.To solve the aforementioned problems,related research on urban public transit OD demand estimation method based on multi-source data fusion is carried out in this paper.Firstly,the commonly used bus IC data in urban public transit system is taken as the main data source,and is combined with the bus GPS data and bus line stop data to deduct the OD demand.The collection,transmition and store process of these data are discussed,and the corresponding data preprocessing is carried out,which mainly includes the the selection of research period,the screening of key fields and the cleaning of invalid data.On the board site identification algorithm,the bus line stop data and bus GPS data are integrated to obtain the time information of bus when arriving a specific site to form a bus arrival timetable.The algorithm based on density clustering is used to correct the basic geolocation coordinate system of bus GPS data and bus line stop data.By matching the bus arrival timetable with bus IC data,the board site can be identified.A system time deviation of bus arrival timetable and bus IC data is also considered by introducing the hysteresis coefficient for time correction due to passenger's non-instantaneous when swiping cards on bording.On the depart site identification algorithm,the bus travel chain theory is adopted to analyze the bus travel process.For bus IC data that satisfies the bus travel chain theory,the depart sites are obtained by directly use bus travel chain theory.For those not satisfied with bus travel chain theory,it is assumed that it obeys the same data distribution as those satisfied with bus travel chain theory.A deep feedforward network classification model is trained with the input of depart sites of bus IC data following bus travel chain theory as priori knowledge.The bus IC data which are not satisfied with bus travel chain theory are then used as “to be predicted” part and are predicted using the pretrained deep feedforward network classification model to get the depart sites.Finally,the Guiyang bus IC data,bus GPS data and bus line stop data are taken as an example to verify the effectiveness and applicability of the proposed algorithm.