Analysis Of Transit Ridership Structural Changes Before And After The Opening Of The Metro Based On Large Sample Data

In recent years,the construction of rail transit has been promoted rapidly.However,at the same time,bus ridership in large cities has generally declined rapidly,and the overall proportion of public transport in motorized travel has increased slowly.This has led to a series of related puzzles and disputes.In this context,the empirical research on the changes in transit passenger flow before and after the opening of the metro in cities with millions of people is of great value.Although various big data resources provide technical conditions for the analysis of public transit ridership structure after the metro is put into operation,how to extract the variation characteristics of spatiotemporal distribution and travel patterns is an urgent technical problem to be solved.Therefore,this paper uses the multi-source public transport data of Xiamen Metro Line 1 for empirical research.This research proposes to extract the spatial and temporal characteristics,user activity patterns,and bus ridership structural changes in the public transport network to establish a more in-depth and comprehensive technical method to describe the changes before and after the metro opening from the macro transit network level,the middle bus line-station level and the micro individual level.Specifically,the main research work of this paper can be summarized as the following four points:(1)Preprocessing of multi-source transit large sample data and OD constructionThere are many original tables of multi-source transit large sample data.Moreover,there are missing,errors,one to many,and other phenomena in the connection relationship.The selection of connection paths and key fields has a great impact on preprocessing results of an information redundancy environment.To measure the connection effect of each path,theoretical matching rates and actual matching rates were defined.The connection relationship was the only factor taken into account in the theoretical matching rate without actual calculation,while connection relationship,data loss,algorithm error,and other factors were taken into account in the actual matching rate.Differences between the two matching rates were used to analyze the reasons for connection errors.The method of allocating the operation time according to the proportion of station spacing was used to correct the GPS data missing and reduce ridership difference caused by data quality problems before and after the metro opening of the subway.Bus alighting stations and time of individual-level were inferred.A Multi-source OD construction was built after the definition of transfer and determination of threshold considering duration and distance of activities.(2)Spatiotemporal heterogeneity of macro traffic demandLarge sample data resources of public transport make it possible to observe continuously for many days.Simply applying traditional methods can not make full use of its information resources.The method of matrix dimensionality reduction can extract the spatiotemporal features of the OD matrix.The Singular Value Decomposition(SVD)matrix dimensionality reduction algorithm was used to analyze the temporal and spatial characteristics of the OD spatiotemporal matrix at the station level.The Go Decomposition(Go Dec)matrix dimensionality reduction algorithm was used to analyze the law of stable passenger flow and disturbed passenger flow at the station group level.The opening of the metro has had a continuous impact on the BRT corridor.The rank-sum test results of the simplest row of the low-rank matrix before and after the metro opening show that the global public transit structure changes have not occurred soon.It needs to be tracked and evaluated continuously.(3)Travel variability of micro individualFor the passive large sample public transport data without socio-economic attributes and travel purposes,the Density-Based Spatial Clustering of Applications with Noise(DBSCAN)algorithm was used to represent the habitual travel with core points and boundary points,and the random travel with noise points,to increase the recognition depth of user behavior patterns.Differences between habitual travel and random travel were observed from three aspects of statistical characteristics,temporal distribution,and spatial distribution.Transfer or not and the transfer degree were defined by the analogy of the small sample questionnaire survey.Taking the combination of(passenger-habitual OD)as the primary key,passengers who were transferred to the metro subsystem and attracted to the metro subsystem were defined.These passengers were measured from the main mode,transfer,spatial relationship with metro,cross-sea,travel time,and distance.The proportion of passengers who were transferred to the metro subsystem is3.52%,while the proportion of passengers who were attracted to the metro subsystem is 0.83%.A large number of passengers transferring to the metro do not take the metro as the main metro mode in their habitual travel,but take the metro as an attempt.Nearly two-thirds of the passengers attracted to the metro take the subway as the main transit mode,and do not use other transit modes except the metro in their habitual travel.(4)Bus ridership structural change and related factorsThe Iterative Cumulative Sums of Squares(ICSS)algorithm was used to detect structural changes in daily ridership time series from temporal and spatial dimensions when there were major changes in the public transport network.The Synthetic Control Method(SCM)was used to verify the rationality of structural change point identification results and measure impact.10 variables were used to cluster from three aspects: network topology,ridership structural characteristics,and land use,to explore the causes of structural change points.After the metro opening,the passengers along the metro may transfer from bus to metro,resulting in a decline in bus ridership.Highfrequency habitual passengers who are not in the subway corridor may increase transfer or induce bus travel,resulting in a rise in bus ridership.It emphasizes the importance of the space with a high proportion of habitual travel not in the metro corridor in the small sample questionnaire survey and route optimization and adjustment.The main contributions of this paper are as follows:· In terms of data preprocessing,a data fusion preprocessing and OD inference method is proposed for the integrated public systems including three subsystems.· At the level of the macro transit network,this paper presents a method to separate and extract spatiotemporal variation features from transit time-varying OD matrix with a certain observation duration.· At the middle line-station level,in the time series of bus ridership,a technical method is proposed to automatically identify the structural changes in line and station passenger flow.· At the micro individual level,aiming at the public transport data that can track individual electronic footprints,a method for distinguishing and identifying individual spatial activity pattern attributes is proposed.An individual-level definition of mode transfer and a method to measure the degree of mode transfer are proposed.