Research On Key Stations Identification Method In Urban Rail Transit Network Based On Data-driven

With the rapid advancement of the urban rail transit network in our country,the rail transit has gradually developed into the preferred mode of transportation for passengers.However,due to the large passenger flow during the peak period,the passenger flow of the road network will be congested,which will not only cause the reduction of the passenger service level in the road network,but also bring certain difficulties to the improvement of the level of subway passenger flow organization.In order to formulate more targeted passenger transportation organization measures to improve the level of passenger transportation organization at stations,this paper takes the identification of key stations in the urban rail transit network during the morning rush hour as the research objective.Based on the historical passenger flow data,the future road network and station status are accurately predicted.The congested periods and normalized key stations are further evaluated.Finally,the Beijing rail transit network is selected for example analysis.The main work contents of this paper are as follows:(1)The relevant theories of key station identification in urban rail transit network are summarized.Firstly,the concept of key stations in the rail transit network is summarized based on the analysis of passenger traffic congestion in the network.Secondly,according to the nature of the crowded passenger flow in the station,the key stations are classified.Then the causes of the key stations are analyzed and the static and dynamic characteristics of the key stations are summarized.Finally,the practical application based on the identification results of key stations is described.This has laid a theoretical foundation for the follow-up research.(2)A prediction method for the state of urban rail transit network and station passenger flow is proposed.Firstly,the road network and station status evaluation indicators are established.Secondly,based on the AFC and road network topology data,a prediction model combining the Graph Convolutional Network and the Gated Recurrent Unit(TGCN)is proposed to predict the interval full load rate and the number of passengers in the station.Then the evaluation index of the road network and the predicted state of the station is calculated.This provides data support for key station identification.(3)A method for identifying key stations in the urban rail transit network is proposed.The identification of key stations specifically includes the classification of the historical state of the road network and stations,the evaluation of the predicted state of the road network and stations,and the analysis of the relevance of key stations.Firstly,the DBSCAN improved Gaussian Mixture Model Clustering is used to classify the historical state of the road network and station,to realize the automatic identification of the optimal number of classifications and the evaluation of the classification effect.Secondly,the K Nearest Neighbor Classification(KNN)algorithm is used to identify and determine the future road network state level.According to the state level of the road network,the congested periods are distinguished and the key stations are identified in the congested periods.Finally,based on the Origin and Destination(OD)volume,the correlation degree of the key station is analyzed and the key station area is divided.(4)Taking the rail transit network in Beijing as an example,the key station identification method and theory proposed in this paper are verified.The results are as follows: 1)The TGCN model with comprehensive temporal and spatial characteristics proposed in this paper can accurately predict the state of rail transit road network and stations,and the prediction performance is improved to a certain extent than the baseline model.2)This paper proposes an improved Gaussian mixture model clustering classification method based on DBSCAN,and constructs the analysis and calculation of the temporal and spatial correlation of key stations in the road network,which solves the problems of mining massive data and the internal relationship of the point-line network state.The methods proposed in this paper can prospectively identify the key stations in the rail transit road network during the morning peak congestion period,which can provide theoretical and technical support for operators to develop coordinated passenger organization measures in multiple key stations.There are 46 figures,12 tables and 63 references.