Table of Contents Author Guidelines Submit a Manuscript
Discrete Dynamics in Nature and Society
Volume 2016, Article ID 9865230, 10 pages
http://dx.doi.org/10.1155/2016/9865230
Research Article

Weighted Complex Network Analysis of the Different Patterns of Metro Traffic Flows on Weekday and Weekend

1School of Traffic & Transportation Engineering, Central South University, Changsha 410075, China
2MOE Key Laboratory for Urban Transportation Complex Systems Theory and Technology, Beijing Jiaotong University, Beijing 100044, China

Received 15 October 2016; Accepted 15 November 2016

Academic Editor: Ricardo López-Ruiz

Copyright © 2016 Jia Feng et al. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

Abstract

We present a multilayer model to characterize the weekday and weekend patterns in terms of the spatiotemporal flow size distributions in subway networks, based on trip data and operation timetables obtained from the Beijing Subway System. We also investigate the disparity of incoming and outgoing flows at a given station to describe the different spatial structure performance between transfer and nontransfer stations. In addition, we describe the essential interactions between PFN and TFN by defining an indicator, real load. By comparing with the two patterns on weekday and weekend, we found that the substantial trends have roughly the same form, with noticeable lower sizes of flows on weekend ascribed to the essential characteristics of travel demand.

1. Introduction

In recent years, complex network theory has become an important approach to the study of the structure and dynamics of traffic networks. Mining spatiotemporal statistical regularities of the human behavior is a common focus of statistical physics and complexity sciences. The human behavior spatiotemporal regularities are the bases of reasonable traffic infrastructure planning, improvement of level of service, and traffic jam control. The patterns on weekday and weekend have been paid more attentions because they affect common lives. Because subway is an important subsystem of urban commuting transportation system, the travel demands on weekdays are quite different from that on weekends.

Transportation engineers studied the differences of human travel behaviors between weekdays and weekends by collecting databases with the survey data [17]. However, some problems are still observed in the survey data approach. For example, survey data is disadvantaged by high cost, low frequency, and small sample size [7]. Moreover, it still lacks a precise method to analyze the statistical regularities of a single mean of transportation.

Nowadays, with the development of electronic technique, more and more approaches (such as mobile phone data, GPS, and smart card) can be used for recording the data of human individual spatiotemporal movements. The data provide a possibility to analyze human travel behavior statistically [813]. Complex network theory is a research field that is concerned with the connections and interactions among components in a system, and it has offered an important approach to study the structure of subway systems [1420]. Recently, studies of metro traffic flow distributions underlying the physical topology with a weighted network-based approach have proved that complex network is a useful method for human travel behavior statistical analysis [2127]. Kurant and Thiran [21, 22] were the first researchers to analyze train flow networks using timetables from Warsaw’s mass transportation system; they found the edge weight and node strength distributions to be heavily right-skewed and heterogeneous, although they did not focus on the differences between weekdays train flow patterns and those on weekends. Soh et al. [24] applied a complex weighted network to the passenger flows of the Singapore Rapid Transit System (RTS) and concluded that weighted eigenvector centralities elucidated significant differences in the passenger flows on weekdays and weekends, although they use an average weight to describe passenger moving between the different nodes on weekdays and weekend. But they have focused only on passenger flow networks. However, because traffic data is difficult to collect, previous studies have usually focused on the physical topology of subway systems, whereas few studies have considered the differences between traffic flow characteristics on weekdays and weekends.

First, this paper aims to obtain statistical properties, such as weights and strengths distributions, that characterize the structure and behavior of both passenger flow network (PFN) and train flow network (TFN) on weekday (WDPFN, WDTFN) and weekend (WEPFN, WETFN). Second, we suggest appropriate ways to measure the essential interactions between patterns in PFN and TFN and to explain the meaning of these results moreover. Third, we give helpful suggestion to improve the organizations of metro systems.

A comparison between our work and the existing literature considering travel behaviors on weekdays and weekends is given by Table 1.

Table 1: Parameters used in studies of different travel behaviors on weekdays and weekends.

This paper is organized as follows. In Section 2 we present a multilayer model to analyze traffic flow patterns in subway networks based on trip data and operation timetables obtained from the Beijing Subway System. Section 3 presents the weekday and weekend patterns and the differences between these two patterns. Section 4 summarizes the results.

2. Data Preparation

2.1. Graph Dataset

In this study, we define a graph of a subway network with stations connected by bi-directional edges. Our data contains the stations with their physical coordinates in Beijing Subway System (BSS). As of May, 2015, the BSS network has 18 lines; unique stations and sections between stations in operation, which can be seen in Figure 1.

Figure 1: The subway real physical network of the BSS. The metro network and the road network are presented by white and gray lines, respectively. Moreover, the five lines we studied in this paper are shown by color.
2.2. Smart Card Data

The smart card system used in BSS is called Yikatong electronic ticketing. The Yikatong electronic ticketing dataset contains precise timing and location information for both boarding and alighting.

This present study is conducted based on smart card records on Tuesday, April 15, 2014, and Saturday, April 19, 2014.

2.3. Timetables

BSS operation timetable describes the arrival and departure times for each station at which the trains stop, the length of time that each train stays at each station, and the running times for each track section.

Our train flow analysis uses BSS operation timetable from the same dates. It is important to note there are two different timetables on weekdays and weekends.

2.4. Case Study

Because the data is difficulty to collect, our study focuses on a subsystem network that consists of 5 lines; unique stations, and sections.

Moreover, the weekday datasets encompassed a total of 44,424 trains and 27,197,333 individual passengers moving through the network on Tuesday, April 15, 2014. And the weekend datasets encompassed a total of 42,965 trains and 7,060,927 individual passengers.

A train flow matrix and a passenger flow matrix were constructed to analyze the BSS train and passenger trip data. The elements of and , respectively, represented the numbers of trains and passengers taking trips between a pair of adjacent stations over a given period time. All of the data was divided into half-hour segments, so that the given period  min was the time interval over which the numbers of passenger movements were aggregated. The analysis of these flow matrices over several time intervals at different operational times of day will explicitly define the traffic flow patterns within the network.

3. Model

We constructed two directed weighted networks TFN and PFN of a subway system by incorporating the trip data and the timetables. Figure 2 presents a simplified example of how the passenger transport demand through an operation service network (train flow network) and finally forms a passenger flow network on a physical network of nodes and directed edges. Sometimes the passenger demand cannot be satisfied by the train capacity; bottleneck occurs in the subway system. Thus, the passenger flow network is not mapped directly by the passenger transport demand network when the bottleneck occurs.

Figure 2: Schematic representation of an example of (a) the passenger transport demand of the subway infrastructure network with nodes and edges, (b) the train flow network with nodes and edges, and (c) the corresponding passenger flow network with nodes and edges.

As shown in Figure 2, there are a passenger demand network and two traffic flow networks PFN and TFN, which are represented as three directed weighted graph , , and .

3.1. Weight and Strength

In order to have a preliminary grasp on the data, we firstly analyze the weight and strength of traffic flow. An associated weighted adjacency matrix representing the train flow from station to station and an associated weighted adjacency matrix representing the passenger flow from station to station as follows

The above subscripts and appear as destination-source station; for example, denotes the train flow of departure (outflow) of station , moving in the direction, while denotes the train flow of arrival (inflow) of station , moving in the direction. It is noted that in our content the subscripts particularly refer to a pair of two adjacent stations unless otherwise mentioned.

Subsequently, in TFN, a node’s strength of station noted as is simply denoted as the sum of the weights on the edge incident upon it and shall be defined as

In PFN, the node strength of station noted as shall be defined as

Thus, we define the traffic flows on weekday and weekend, respectively, as and and and and the strength as and and and .

Furthermore, in PFN, there are considerable passenger flows moving into station from outside the network and moving out of the network for the same station, which can be denoted by and as follows

We define the incoming flow and outgoing flow at a given station in order to explore the average crowdedness and utilization rate of a station.

3.2. Disparity

We now define the time averaged incoming passenger flow and outgoing flow at a given station:

And we define that the disparity [28, 29] in the weights of a given station can be evaluated by the quantities and defined as [30, 31]

3.3. Real Load

The real load [21, 22] of a node is the sum of the weights of all logical edges whose paths traverse this node. In a subway system, there are three kinds of real load: , , and . The first two are the train flows and passenger flows loading on the physical topology of subway system network, which can be presented by node strength. However, the third one is to measure the passenger flows average crowdedness underlying the TFN, which fully reveals the essential interactions between PFN and TFN.

An associated weighted adjacency matrix representing the passenger flow underlying the train flow network from station to station j:

4. Results and Discussions

4.1. Weight and Strength Distributions

In order to get a preliminary grasp on the data, we first obtain statistical properties of the train flow and passenger flow networks for the BSS. The basic statistical properties of weight and strength distributions are listed in Table 2. According to Table 2, the sizes and fitted distributions of PFN and TFN are obviously different. In order to get a deeper insight of these two networks, we carried out a weighted analysis and plotted the results, respectively, in Figures 3 and 4. Figure 3 shows the distributions of the number of passenger trips between two adjacent stations and the distributions of the number of passengers one station handles, respectively, on weekday and weekend.

Table 2: Statistical properties of the PFN and TFN on weekday and weekend.
Figure 3: In PFN, flow weight and node strength distributions. The weights of passenger flow and can be fitted by power law distributions with exponents and , respectively. The goodness-of-fit () are all 0.99. And the strengths and can be fitted by power law distributions with exponents and , respectively. The goodness-of-fit () are 0.99 and 0.56.
Figure 4: In TFN, flow weight and node strength distributions. The weights of train flow and can be fitted by Weibull distributions with parameters and and and , respectively. The goodness-of-fit () are 0.77 and 0.93. And the strengths and can be fitted by Weibull distributions with parameters and and and , respectively. The goodness-of-fit () are 0.90 and 0.90.

There are three aspects that have to be addressed and can be seen in Figure 3. Firstly, it is observed that all of the passenger flow weight and strength distributions and are significant right-skilled on log-log scale which can be fitted by power law distributions. This indicates that the passenger flow patterns vary in intensity, and there exist travel routes and hub nodes with very high traffic. This kind of heterogeneous passenger flow organization is corroborated by previous studies [23, 24, 27]. However, compared to previous studies, our case networks are more significant than the Metropolitan Seoul Subway system () [23] and more noticeable than the Singapore Rapid Transit system ( on weekday and on weekend) [24] but are similar to the total system of BSS () [27]. This can be attributed to the heterogeneous flow intensities caused by an urban spatial mismatch. Secondly, the sizes of passenger flows on weekend are considerably lower than those on weekday, especially at the travel routes and hub nodes handling high traffic. This indicates that the travel demand on weekend is lower than that on weekday. This discrepancy in magnitude shown here implies different travel patterns between weekdays and weekends, as people would have to choose different way to fulfill their goals (such as shopping, socializing, entertainment, and leisure) on weekend, which has been reported in human travel activities survey studies [3, 5, 6]. For example, the average number of daily person trips for every traveler is 3.40 for weekdays and 3.14 and 2.85 for Saturdays and Sundays [6]. Thirdly, the weight and strength distribution lines on weekend both have a significant turning point in the plots, which are not as smooth as those on weekday.

As shown in Figure 4, the train flow weight distributions have significantly unimodal trend with a skewness leaned to the front and a smoothly light tail on log-log scales, which can be fitted by Weibull distributions. The strength distributions present a clear right-skewed asymmetry trend and a heavy-tailed characteristic on log-log scales, which also can be fitted by Weibull distributions. It is clear that the sizes of train flow on weekend are slightly lower than those on weekday. It is important to note that there are two different timetables on weekday and weekend in order to match the working and nonworking day passenger flow patterns. For the total passenger trip data on weekend is an order of magnitude lower than the data on weekday and the operational train numbers have been reduced. Moreover, the form of fitted distributions in TFN is the Weibull distribution, which is different from that in PFN. The fundamental differences between these two coexisting flow networks may be attributed to the fact that the dynamic evolution rules of TFN are similar to a branching process and the dynamic patterns of PFN are self-organized. In other words, in TFN the size of newly train flow is determined by the redistribution of the size of preexisting flow intensities; thus the Weibull distribution is expected to emerge [25, 32]. On the contrary, in PFN the size of newly passenger flow is determined by not only the size of preexisting flow but also the passenger flows moving between that station and outside of the network, which will be studied in Section 4.2

4.2. Disparity Distributions in PFN

Now we aim to find and highlight the patterns of passenger flows both taking trips between adjacent stations and using a station (i.e., entering into and leaving a station). The size distributions of total incoming flow and outgoing flow at a given station are presented in Figure 5.

Figure 5: Incoming and outgoing passenger flows distributions. The weight of incoming flow on weekday and weekend and can be fitted by exponential distributions with exponents and , respectively. The goodness-of-fit () are 0.97 and 0.96. The weight of outgoing flows and can be fitted by exponential distributions with exponents and , respectively. The goodness-of-fit () are both 0.97.

As detailed in Figure 5, the weight distribution patterns of incoming and outgoing passenger flows show mostly noticeable declined characteristics, smoothly in the middle and rapidly at the front and the tail. First, the incoming and outgoing flows on both weekday and weekend can be fitted by exponential distributions. This pattern displays the same trend as the cases observed in many works, such as metro passenger flows [27] and Internet traffic [31]. And this exponential trend means that the spatial structure of total incoming and outgoing flows in PFN display a bilevel performance [31]. Second, it is not a surprise to find that the weights on weekday are both larger than those on weekend, which is attributed to the travel demand. Roughly, the orders of most visited and passenger sending stations are stable on both weekday and weekend. To clearly characterize the fine structure of the incoming and outgoing flows at station i, we plot the disparity distribution in Figure 6.

Figure 6: The probability distributions of for both incoming and outgoing flows on (a) weekday and (b) weekend. The distributions on weekday are peaked for ≈ 0.45 and ≈ 0.24 and both concentrated in the range (0.20, 0.25) and (0.30, 0.48). The distributions on weekend are peaked for ≈ 0.45 and ≈ 0.43 and both concentrated in the range (0.20, 0.26) and (0.30, 0.50).

There are three aspects that have to be noticed in Figure 6. The first question involves that the distributions all present a special bimodal structure. This is a further illustration of the bilevel performance presented in spatial patterns structure. The second problem relates to the differences between and that the are larger in small value range and are larger in large value range. In PFN, a small value of means that the weights on each adjacent edge are similar, and a large value indicates the existence of dominative edge. Thus, the spatial patterns of incoming flows are scattered and those of outgoing flows seem to be more concentrated. The third aspect deals with the fact that the values of are similar to ( is the number of adjective stations of station ) 1/2 = 0.5 (except for few transfer stations with more than 2 adjective stations and terminal stations with only 1 adjective station). This high homogeneity means that the weights of incoming and outgoing flows in each level are relatively close.

4.3. Real Load Distribution

Next we investigate the real load distributions of the two network systems: WD and WE. It is important to note that the real load distribution here is an important indicator to analyze the correlation between PFN and TFN. This indicator can be seen as appropriate measure of the train capacity utilization rate in the metro system, which can be interpreted as the passenger density in a train. Figure 7 shows the distributions and the temporal patterns of the real load on weekday and weekend. In order to get a deeper insight into the structure of the two real load networks, we present the spatial patterns in Figure 8.

Figure 7: Real load (a) distributions on log-log scale and temporal patterns of (b) weekday and (c) weekend. In (a), the real loads and can be fitted by power law fits with exponents 0.55 and 0.56, respectively. The are all 0.97.
Figure 8: Real load spatial patterns on weekday and weekend.

In Figure 7(a), we can clearly observe that the real load distributions and are remarkable right-skilled on log-log scale which can be fitted by power law distributions and show markedly heavy-tailed characteristics. Moreover, the quantities of are lower than in the tail part of distribution line, indicating that the real load on weekend is less than that on weekday. Figure 7(b) presents the temporal patterns of passenger traveling in TFN on weekday. It is clear that there are two distinct peaks—07:00~10:00 and 18:00~20:00 corresponding to the peak hours on working day. Moreover, the morning peak is more prominent than the evening peak in both the traffic density and the temporal duration. As apparent from Figure 7(c), there are two moderate morning and evening peaks, 07:00~09:00 and 19:00~22:00 with the morning peak more prominent in traffic density and the evening peak more prominent in the temporal duration. Comparing Figures 7(b) and 7(c), the peaks on weekend are much lower than those on weekday, which indicates a highly uneven utilization of the underlying train flow network on weekday and a highly level of service supplied by the metro system on weekend. It could be concluded that the temporal pattern on weekday and weekend networks presents a dissimilar match as expected.

As illustrated by Figure 8(a), the spatial clustering of a large number of heavy-burden nodes is distributed along Line 1’s east-west route and Line 5’s north-south route on weekday. On the other hand, the spatial pattern on weekend presents the clustering phenomenon around the transfer stations. In summary, the dense distribution areas in the spatiotemporal pattern indicate bottlenecks in the train capacity, which can aid in the effective operation of train services.

In this paper, we found and highlighted the statistical properties and spatiotemporal patterns including the size distributions in PFN that can be well approximated with a power law distribution which indicates that the passenger flow patterns vary in intensity with hub nodes and busy edges on both weekday and weekend; by contrast, the fitted form of the size distributions in TFN is Weibull, which may be attributed to the fact that the dynamic evolution in TFN is similar to a branching process for there are no train flows exchanged between TFN and outside. We also suggest using a quantity to measure the disparity of each node in PFN, and the results show a significant two-level performance caused by the different dynamics flows evolution process between the transfer and nontransfer stations.

By comparing with the two patterns on weekday and weekend, we found that the substantial trends have the roughly same form, with noticeable lower sizes of flows on weekend ascribed to the essential characteristics of travel demand on weekends. As a result, the real load on weekend presents a loose and comfortable performance on spatiotemporal patterns.

5. Conclusions

Prior work has documented that complex network methods are useful for studying the underlying physical topology of traffic flows in metro systems. However, little research has been conducted on both passenger and train flow patterns or intrinsic differences between weekday and weekend.

In this paper, we found and highlighted the statistical properties and spatiotemporal patterns in PFN and TFN based on a contribution of two directed weighted networks using the trip dataset collected from smart card transactions and the train movement dataset processed from operation timetable. We characterized the flow weight and strength distributions and found that the heterogeneous feature of passenger flow shows a self-organized pattern in which spatial mismatch occurs as individual passengers commute through the city’s center; on the other hand, the train flows evolution dynamics present a redistribution of preexisting flow intensities. We also investigated the disparity of incoming and outgoing flows at a given station to describe the bilevel performance (transfer and nontransfer stations) of the spatial structure. We further discussed the fundamental correlation between PFN and TFN by defining an indicator, real load, and characterizing the size distribution and spatiotemporal patterns on weekday and weekend, respectively. By comparing with the two different patterns, we found that the real load on weekend presents a loose and comfortable performance in whole temporal duration and that on weekday shows two obvious congested periods of time, which are the results of the sizes of flows on weekend that are noticeably lower than those on weekday ascribed to the essential characteristics of travel demand on weekends.

The empirical findings can give us some useful insights on patterns of urban human mobility within a large metro network on both weekday and weekend. As travel demand increases and infrastructure construction is constrained, traffic congestion takes place not only at morning and evening peak hours on weekdays but also in major shopping centers, sports arenas, and recreational areas in big cities over weekends. Adopting a complex network approach to study the passenger and train flow patterns can therefore be beneficial for effective operation of train services.

This study is limited in that only five subway lines were analyzed because of the difficulty associated with data collection, and research on the overall system with more lines should be conducted to further validate the above conclusions. Moreover, other traffic flow models of subway networks could be explored from a variety of layered perspectives.

Competing Interests

The authors declare that they have no competing interests.

Acknowledgments

The research described in this paper was substantially supported by the National Natural Science Foundation of China, Project U1334207, the China Postdoctoral Science Foundation, Project 2015M582347, and The Postdoctoral Science Foundation of Central South University.

References

  1. H. J. Kim, D. H. Kim, and J.-H. Chung, “Weekend activity and travel behavior in a developing country: empirical study using multilevel structural equation models,” Transportation Research Record, vol. 1894, no. 1, pp. 99–108, 2004. View at Google Scholar · View at Scopus
  2. H. Verreault and C. Morency, “Transcending the typical weekday with large-scale single-day survey samples,” Transportation Research Record: Journal of the Transportation Research Board, vol. 2230, pp. 38–47, 2011. View at Google Scholar
  3. M. Zhong, D. J. Hunt, and X. Lu, “Studying differences of household weekday and weekend activities,” Transportation Research Record: Journal of the Transportation Research Board, vol. 2054, pp. 28–36, 2008. View at Publisher · View at Google Scholar
  4. K. Born, S. Yasmin, D. You, N. Eluru, C. Bhat, and R. Pendyala, “Joint model of weekend discretionary activity participation and episode duration,” Transportation Research Record, no. 2413, pp. 34–44, 2014. View at Publisher · View at Google Scholar · View at Scopus
  5. A. Aguiléra, M.-H. Massot, and L. Proulhac, “Exploring the relationship between work and travel behavior on weekdays,” Transportation Research Record, no. 2135, pp. 69–77, 2009. View at Publisher · View at Google Scholar · View at Scopus
  6. A. M. Lockwood, S. Srinivasan, and C. R. Bhat, “Exploratory analysis of weekend activity patterns in the San Francisco Bay Area, California,” Transportation Research Record, no. 1926, pp. 70–78, 2005. View at Google Scholar · View at Scopus
  7. S. Jiang, J. Ferreira, and M. C. González, “Clustering daily patterns of human activities in the city,” Data Mining and Knowledge Discovery, vol. 25, no. 3, pp. 478–510, 2012. View at Publisher · View at Google Scholar · View at Scopus
  8. X. Huang and J. Tan, “Understanding spatio-temporal mobility patterns for seniors, child/student and adult using smart card data,” The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, vol. XL-1, pp. 167–172, 2014. View at Google Scholar
  9. A.-S. Briand, E. Côme, M. K. El Mahrsi, and L. Oukhellou, “A mixture model clustering approach for temporal passenger pattern characterization in public transport,” International Journal of Data Science and Analytics, vol. 1, no. 1, pp. 37–50, 2016. View at Publisher · View at Google Scholar
  10. X. Liang, X. Zheng, W. Lv, T. Zhu, and K. Xu, “The scaling of human mobility by taxis is exponential,” Physica A: Statistical Mechanics and Its Applications, vol. 391, no. 5, pp. 2135–2144, 2012. View at Publisher · View at Google Scholar · View at Scopus
  11. B. Jiang, J. Yin, and S. Zhao, “Characterizing the human mobility pattern in a large street network,” Physical Review E: Statistical Nonlinear & Soft Matter Physics, vol. 80, no. 1, pp. 1711–1715, 2009. View at Google Scholar
  12. C. Peng, X. Jin, K.-C. Wong, M. Shi, and P. Liò, “Collective human mobility pattern from taxi trips in urban area,” PLoS ONE, vol. 7, no. 4, article e34487, 2012. View at Publisher · View at Google Scholar · View at Scopus
  13. C. Roth, S. M. Kang, M. Batty, and M. Barthélemy, “Structure of urban movements: polycentric activity and entangled hierarchical flows,” PLoS ONE, vol. 6, no. 1, Article ID e15923, 2011. View at Publisher · View at Google Scholar · View at Scopus
  14. V. Latora and M. Marchiori, “Is the Boston subway a small-world network?” Physica A: Statistical Mechanics & Its Applications, vol. 314, no. 1–4, pp. 109–113, 2002. View at Publisher · View at Google Scholar · View at Scopus
  15. J. Sienkiewicz and J. A. Hołyst, “Statistical analysis of 22 public transport networks in Poland,” Physical Review E, vol. 72, no. 4, Article ID 046127, 2005. View at Publisher · View at Google Scholar · View at Scopus
  16. P. Angeloudis and D. Fisk, “Large subway systems as complex networks,” Physica A: Statistical Mechanics & Its Applications, vol. 367, pp. 553–558, 2006. View at Publisher · View at Google Scholar · View at Scopus
  17. S. Derrible and C. Kennedy, “The complexity and robustness of metro networks,” Physica A: Statistical Mechanics and Its Applications, vol. 389, no. 17, pp. 3678–3691, 2010. View at Publisher · View at Google Scholar · View at Scopus
  18. J. Lin and Y. Ban, “Complex network topology of transportation systems,” Transport Reviews, vol. 33, no. 6, pp. 658–685, 2013. View at Publisher · View at Google Scholar · View at Scopus
  19. Y. Liu and Y. Tan, “Complexity modeling and stability analysis of urban subway network: Wuhan City Case Study,” Procedia—Social and Behavioral Sciences, vol. 96, pp. 1611–1621, 2013. View at Publisher · View at Google Scholar
  20. B. Gao, Y. Qin, X.-M. Xiao, and L.-X. Zhu, “K-means clustering analysis of key nodes and edges in Beijing subway network,” Journal of Transportation Systems Engineering and Information Technology, vol. 14, no. 3, pp. 207–213, 2014. View at Google Scholar · View at Scopus
  21. M. Kurant and P. Thiran, “Extraction and analysis of traffic and topologies of transportation networks,” Physical Review E—Statistical, Nonlinear, and Soft Matter Physics, vol. 74, no. 3, Article ID 036114, 2006. View at Publisher · View at Google Scholar · View at Scopus
  22. M. Kurant and P. Thiran, “Layered complex networks,” Physical Review Letters, vol. 96, no. 13, Article ID 138701, 2006. View at Publisher · View at Google Scholar · View at Scopus
  23. K. Lee, W.-S. Jung, J. S. Park, and M. Y. Choi, “Statistical analysis of the Metropolitan Seoul Subway System: network structure and passenger flows,” Physica A: Statistical Mechanics and Its Applications, vol. 387, no. 24, pp. 6231–6234, 2008. View at Publisher · View at Google Scholar · View at Scopus
  24. H. Soh, S. Lim, T. Zhang et al., “Weighted complex network analysis of travel routes on the Singapore public transportation system,” Physica A: Statistical Mechanics and Its Applications, vol. 389, no. 24, pp. 5852–5863, 2010. View at Publisher · View at Google Scholar · View at Scopus
  25. K. Lee, S. Goh, J. S. Park, W.-S. Jung, and M. Y. Choi, “Master equation approach to the intra-urban passenger flow and application to the Metropolitan Seoul Subway system,” Journal of Physics A: Mathematical and Theoretical, vol. 44, no. 11, Article ID 115007, 2011. View at Publisher · View at Google Scholar · View at MathSciNet · View at Scopus
  26. M. A. Ramli, P. C. Monterola, G. L. K. Khoon, and T. H. G. Guang, “A method to ascertain rapid transit systems' throughput distribution using network analysis,” Procedia Computer Science, vol. 29, pp. 1621–1630, 2014. View at Publisher · View at Google Scholar
  27. Q. Xu, B. H. Mao, and Y. Bai, “Network structure of subway passenger flows,” Journal of Statistical Mechanics: Theory and Experiment, vol. 2016, no. 3, Article ID 033404, 2016. View at Publisher · View at Google Scholar
  28. B. Derrida and H. Flyvbjerg, “Statistical properties of randomly broken objects and of multivalley structures in disordered systems,” Journal of Physics A: Mathematical and General, vol. 20, no. 15, pp. 5273–5288, 1987. View at Publisher · View at Google Scholar · View at MathSciNet · View at Scopus
  29. S. Boccaletti, V. Latora, Y. Moreno, M. Chavez, and D.-U. Hwang, “Complex networks: structure and dynamics,” Physics Reports, vol. 424, no. 4-5, pp. 175–308, 2006. View at Publisher · View at Google Scholar · View at MathSciNet · View at Scopus
  30. M. Barthélemy, A. Barrat, R. Pastor-Satorras, and A. Vespignani, “Characterization and modeling of weighted networks,” Physica A: Statistical Mechanics and Its Applications, vol. 346, no. 1-2, pp. 34–43, 2004. View at Publisher · View at Google Scholar · View at Scopus
  31. M. Barthélemy, B. Gondran, and E. Guichard, “Spatial structure of the internet traffic,” Physica A: Statistical Mechanics & Its Applications, vol. 319, no. 7, pp. 633–642, 2003. View at Publisher · View at Google Scholar · View at Scopus
  32. S. Goh, H. W. Kwon, M. Y. Choi, and J.-Y. Fortin, “Emergence of skew distributions in controlled growth processes,” Physical Review E, vol. 82, no. 6, Article ID 061115, 2010. View at Publisher · View at Google Scholar · View at Scopus