Study on the structure characteristics of the passenger flow network in Beijing subway:Based on complex network theory

Shuqi Li2016 · DOI

Abstract

The discovery of the basic law in the subway network system is important to promote the coordinated development of traffic management science and traffic science and technology. Based on complex network theory, this paper constructs a weighted complex network model of passenger flow in Beijing subway in order to study the characteristics of it. The network model is analyzed in the following aspects: the centrality of the weighted degree, the weighted clustering coefficient, and the cluster identification. It is found that a small number of sites have a greater degree of centrality and bear a larger passenger capacity; at the same time, the behavior of the passenger flow has a clear order and cluster. Therefore, it provides a scientific basis for the management of passenger flow in Beijing city.

Introduction

Urban planners traditionally focus on the physical topology of transit systems—the static placement of tracks and stations. However, this perspective often fails to capture the dynamic and complex characteristics of actual passenger usage. As urban populations expand against limited resources, understanding the "invisible network" of passenger flow becomes critical. This study applies complex network theory to Beijing subway data to identify structural hubs and behavioral patterns that differ significantly from the physical layout of the tracks.

Methodology

The research constructs a weighted, directed complex network where subway stations serve as nodes and travel behavior (passengers moving from Station A to Station B) defines the edges. The volume of passengers on these routes determines the edge weights. To analyze this system, the study employs several metrics:

  • Degree Centrality ($D_i$): The count of unique station-to-station connections at a site.
  • Weighted Degree Centrality ($W_i$): The total passenger volume passing through a station.
  • Weighted Clustering Coefficient ($C_i^w$): A measure of local connectivity density, adjusted for traffic volume.
  • Community Detection: The Blondel algorithm (2008) is used to identify network clusters based on modularity ($Q$).

Data and Network Construction

The study utilizes passenger flow data from the Beijing Subway in April 2013, which then comprised 15 lines and 185 stations. While the theoretical network is nearly a complete map, the analysis focuses on the "essential characteristics" by filtering for the top 29.83% of routes. These routes account for 80% of total travel behavior, resulting in a refined network of 179 stations and 9,423 routes.

Analysis of Network Characteristics

The Pareto Principle of Commuting

The distribution of traffic across the network is highly uneven, following a Pareto-like distribution. The analysis reveals that a mere 9.92% of routes carry 50% of the total passenger load. This concentration explains why specific "bottleneck" stations can cause systemic delays across the entire city.

The cumulative percentage of passengers vs. the cumulative percentage of bus routes.

The 150-Connection Tipping Point

A striking finding is the non-linear relationship between a station's connectivity (degree centrality) and its traffic volume (weighted degree). The data identifies a critical threshold at 150 connections:

  • Below 150 connections: Each new route adds approximately 966 passengers ($R^2 = 0.9295$).
  • Above 150 connections: Each new route adds approximately 2,574 passengers ($R^2 = 0.7575$).

This 2.66x increase suggests that once a station becomes a major hub, its sensitivity to network changes and its potential for congestion grow exponentially.

The non-linear relationship between degree centrality and central weighted degree, showing the tipping point at 150 connections.

Hub Connectivity and Clustering

The study finds a strong negative correlation between weighted degree centrality and the clustering coefficient. High-traffic hubs tend to connect to many disparate, unrelated areas rather than being part of localized, "cliquey" station neighborhoods. This "anti-social" behavior of hubs is a key structural feature of the passenger flow network.

Results and Discussion

The analysis of centrality distribution shows that 46% of stations maintain between 45 and 145 connections, representing a moderate level of connectivity for the majority of the system.

Community detection identified eight distinct clusters, with five dominant clusters accounting for 94.41% of all stations. The largest cluster contains 44 stations with intense internal traffic contacts. Critical nodes identified as high-pressure anchors include Guo Mao, Dong Zhimen, Xi Zhimen, and Da Wang Road.

Conclusion

While the 2013 data provides a foundational understanding, the Beijing subway's massive expansion since then likely necessitates updated analysis. The findings suggest a policy of "decentralized management," such as building "shunt" stations near existing hubs to alleviate exponential pressure. Future work should integrate real-time data to develop early-warning systems for passenger surges and conduct detailed cost-benefit analyses of structural interventions.

Summary updated

Abstract

The discovery of the basic law in the subway network system is important to promote the coordinated development of traffic management science and traffic science and technology. Based on complex network theory, this paper constructs a weighted complex network model of passenger flow in Beijing subway in order to study the characteristics of it. The network model is analyzed in the following aspects , the centrality of the weighted degree, the weighted clustering coefficient and the cluster identification. It is found that a small number of sites have a greater degree of centrality, and bear a larger passenger capacity; at the same time, the behavior of the passenger flow has a clear order and cluster. Therefore, it provides a scientific basis for the management of passenger flow in Beijing city.
BibTeX
@misc{li2016study,
  author = {Shuqi Li},
  title = {{Study on the structure characteristics of the passenger flow network in Beijing subway:Based on complex network theory}},
  year = {2016},
  doi = {10.2991/iceep-16.2016.66},
}
Figure