Urban Transportation Modes Recognition Based On Mobile Signaling Data

The rise of big data and the Internet has brought about tremendous changes in transportation and travel.The increasing popularity of smart phones and the rapid development of ICT technology provide an opportunity for mobile data to be applied to travel behavior survey.Mobile signaling data has become the new data source for the study of urban residents’ travel behavior due to its large sample size and full coverage feature.A lot of research fruits have been achieved in travel information extraction,OD analysis and commuting behavior analysis by using mobile signaling data.However,few studies focus on further travel mode identification based on mobile signaling data.Therefore,considering the difficulty of multi-source data acquisition in practice,aiming at rapid migration at the same time,this thesis proposes a method for identifying urban travel modes based on mobile phone signaling data set of urban origins and destinations.Through the deep mining of signaling data,the proposed method can identify five common urban travel modes:walking,bicycle,bus,car and subway.Firstly,after pre-processing and analysis of signaling data based on mobile positioning principle,the basic travel theory was summarized,and the basic travel characteristics and present statistical analysis were extracted from mobile signaling data.Based on the spatial and temporal characteristics of base station user flow,a land use information calibration method for base station is established,and the land use information behind the increment of base station user is extracted by clustering algorithm.The study extracts further land use information including typical residential areas,typical working areas,typical transportation hubs and mixed areas by their obvious unique spatial and temporal passenger flow distribution characteristics.Secondly,a recognition method of urban traffic travel mode based on prior knowledge is proposed.Based on the traditional travel research paradigm,the universal prior knowledge of reasonable travel distance,travel time and average travel speed characteristics of each mode is given on the basis of analyzing the traffic characteristic values of common travel modes.Then,the basic recognition model is built by the theory of fuzzy mathematics,and the general modeling and solving method is given.Based on the calibration results of traffic conditions and base station land use information,the calibration method is revised.Thirdly,a comprehensive recognition method of urban travel mode is proposed based on the feature mining of signaling data.The travel characteristics for observation such as travel distance,travel time and travel average speed are extracted.Based on the clustering theory,the travel mode feature clustering model based on signaling data is built.The intra-cluster features and inter-cluster features are analyzed to identify typical travel modes.The cluster with unclear semantic effect of travel mode is supporting recognized by the fuzzy recognition model.The Comprehensive transportation modes recognition model is given on the basis of feature mining of signaling data.Finally,the whole research process is refined and sublimated,and a macro-medium analysis framework of "cluster analysis of large data samples-intra-class and inter-class feature analysis-local fuzzy processing" is proposed.Based on the signaling data of a working day,this thesis identifies the effective travel mode for all-day travels.By comparing the identification results with the actual travel statistics,it is proved that the proposed method can conduct model division between traffic modes between non-motorized travel modes,bus,car and subway.However,the distinction accuracy between walking and bicycle needs further improvements.