Research On Passenger Group Behavior In Subway Transfer Station Based On Gaussian Mixture Model DBSCAN Algorithm

In recent years,the scale of urban rail transit network has been expanding rapidly.As an important node in the whole network,the transfer station is under great pressure of passenger transportation.Its internal space layout and the basic functions are becoming more and more complicated,resulting in the increasingly diversified behavioral characteristics of passenger groups,which have great impacts on its operation efficiency and service level.Therefore,the research on the behavioral characteristics of the passenger groups in transfer station can provide a reliable reference for the operation and management of the station,and has important application value for the optimization of service facilities,the guidance of passenger organization and the evacuation of abnormal passenger flow.With application of indoor positioning data to study the behavioral characteristics of crowds has been becoming more mature,the paper takes use of the passenger positioning data,which is getting from the Metro Positioning System based on WiFi,to realize the real-time analysis of behavioral aggregation and distribution characteristics of passenger groups in real scene of the whole transfer station by combining with data mining theory.The main research contents include the following three parts:(1)The paper establishes the behavioral characteristics mining model of passenger groups in subway transfer station for the first time.It analyzes the behavioral characteristics of passenger groups with the method of site investigation and modeling simulation,and divides the related problem into four important stages on the basis of data mining model,which provides a reliable theoretical support for the paper and a new idea for the related research of passenger flow in subway transfer station.(2)The paper focuses on proposing the innovative data mining algorithm of passenger positioning data in subway transfer station.Basing on the theoretical model of metro positionging system based on WiFi,the basic acquisition principle and process of passenger positioning data has been studied,and also the data analysis and preprocessing are completed.Then combinging with the characteristics of passenger positioning data,the paper proposes a DBSCAN algorithm based on Gaussian Mixture Model to improve disadvantages of the traditional DBSCAN algorithm and describes the core contents in detail,especially explains how to divide the density distribution heterogeneity dataset into different layers with Gaussian Mixture Model.In addition,the open and simulated data are used to compare and analyze the two algorithms,which prove that the improved algorithm has higher clustering accuracy and better clustering effect.(3)The paper realizes the clustering analysis and visualization of the passenger positionging data in subway transfer station.Taking the actual transfer station as an example,the density clustering analysis of the test passenger positioning data obtained at a certain time is carried out by using the two algorithms respectively.The defects and shortcomings of the traditional DBSCAN algorithm clustering results are clarified,and then the core clustering steps of the improved DBSCAN algorithm such as data fitting,parameter estimation,density stratification,partial clustering are described clearly,also the clustering result and visualization effect of the behavioral characteristics of the passenger groups on the platform has been presented.Finally.the practical application value and significance of mining results are fully expounded.