Travel Characteristic Recognition And Behavior Prediction Of Subway Passengers Based On Individual Spatial-temporal Portraits

The construction of urban rail transit network is advancing rapidly,and the number of passengers who choose to travel by urban rail transit has been increasing.In the operation and management of urban rail transit,the real-time O-D(Origin-Destination)passenger flow prediction is of great significance to ensure operational safety,optimize the network structure,and then realize the refined passenger flow management and vehicle dispatching of urban rail transit.As the travel subject,the passengers have significant differences in their socio-economic attributes and travel preferences.And the rapid development of big data technology provides support for the research of urban rail transit passengers travel,which makes it possible to mine the travel spatial-temporal characteristics of passengers,achieve the refined passenger classification,effectively identify and scientifically predict passenger behaviors from the perspective of individual activities.Therefore,this paper conducts research on the individual travel characteristic identification and travel behavior prediction of urban rail transit passengers based on the analysis of multi-source big data.The main work is as follows:Firstly,analyze and mine the passenger travel AFC data and site GIS information of urban rail transit,establish data cleaning rules,and complete basic travel information preconditioning.Then,an DBSCAN-based method is proposed to extract the individual passenger travel spatial-temporal features,and the passenger travel spatial-temporal feature matrixes are constructed.Based on the quantitatively extracted passenger travel spatial-temporal eigenvalues,the individual passenger travel spatial-temporal portraits is drawn by using HTML5-ECHARTS tool.Taking the passenger travel data of some lines of Suzhou Metro and the whole network of Nanjing Metro as the research objects separately,example analyses are carried out.Secondly,identify the passenger travel characteristics of urban rail transit based on multi-source big data: extract the passenger travel characteristic indexes based on AFC data and site GIS information according to individual travel spatial-temporal portraits;apply the hybrid Poisson distribution model to extract the passenger travel station type index based on the features of land use and passenger flow;extract the passenger travel preference index based on individual economic attributes with the help of crawler tools.And then,taking the relevant data sets of Suzhou Metro and Nanjing Metro as examples,the multidimensional travel characteristic indexes of corresponding passengers are extracted respectively.Thirdly,considering the multi-dimensional travel characteristic index sets of passengers,an unsupervised hierarchical clustering model of individual passengers is constructed based on the two-step clustering-FCM fusion algorithm.Furthermore,an optimization algorithm of passenger classification results based on K-Fold and RCCNR is proposed to further modify the passenger sets in various categories,and realize the refined classification of travel passengers.Taking Suzhou Metro and Nanjing Metro as examples,the model and algorithms are verified and analyzed.Finally,according to the matching rules of individual travel spatial-temporal features,the direct matching method and the Monte Carlo simulation method are used to predict the travel destinations of passengers with fixed regular travel separately.Based on the multi-dimensional travel characteristic index sets of individual passengers,the random forest algorithm and Naive Bayes method are applied to predict the travel destinations of passengers with irregular travel respectively.And the effectiveness of the passenger travel behavior prediction methods based on individual spatial-temporal portraits and multi-dimensional individual travel characteristics is verified by the empirical analyses of the travel data of Suzhou Metro and Nanjing Metro.