City Bike Network Analysis for Bike Sharing System

R. Nagha Akshayaa1 , K Dhipika2, Keerthana V3, Krithika R4

1,2,3,4
Vellore Institute of Technology, Chennai, India
E-mail: [email protected]; [email protected];
[email protected]; [email protected]

R. NAGHA AKSHAYAA 20BCE1325

KRITHIKA R 20BCE1326
DHIPIKA K 20BCE1340
KEERTHANA R 20BCE1561

Abstract

Bike-sharing systems (BSSs) not only facilitate people with the trouble of “the first and last
mile” but also provide them with a sustainable and carbon-free mode of transportation. In
1965, an NGO called Provo established a public bicycle system (PBS) to reduce air pollution
and relieve traffic congestion in Amsterdam, which is regarded as the prototype of BSSs. To
know how people use it on a daily basis is a fundamental resource to plan future expansions
and to make the most out of the public investments.

For the past decades, transportation systems are commonly analyzed as networks. This
abstraction allows to reduce the amount of information available and focus primarily on the
internal structure of the underlying systems. This project analyses the bike-sharing system of
urban cities by working on open dataset of “Helsinki City Bikes”. Here, we present an
analysis of the bicycle-sharing system using networks.

Literature Review

In the study performed by authors in [1], taxi trajectory data from Chengdu and New York
City was analyzed using complex network theory to gain insight into urban structure and
human mobility. The urban trip networks are constructed and analyzed in terms of low-order
organization (degree distribution, cluster-degree coefficient, rich-club coefficient), meso-
order organization (community detection), and higher-order organization (critical nodes and
regions). The goal is to understand the relationship between human mobility and urban
characteristics, as well as to potentially improve urban planning. The findings help to better
understand the urban structure and human mobility. The results show a relationship between
network density and trip distance, heterogeneity in the Chengdu urban trip network,
mismatched community boundaries with administrative boundaries, and critical nodes and
regions in the urban trip networks. These findings have potential applications for urban
planning.

This paper [2] focuses on bike-sharing systems (BSSs) and the studies that have been
conducted on them. The authors note that despite the benefits of BSSs in terms of reducing air
pollution and traffic congestion, issues such as untimely rebalancing operations and
unreasonable station distributions have led to a decrease in user satisfaction and an increase in
maintenance costs. They also mention that many studies have been done on the factors that
affect the usage of bike-sharing, such as the characteristics of users and stations, weather
conditions, and relationship with other means of transportation. However, there is a lack of
research on the relationship between the internal stations of BSSs. To address this gap, the
authors aim to build public bicycle networks using Gephi software to analyze the internal
correlation characteristics of BSSs. In conclusion, by analyzing the relationship between
stations in a bike-sharing system, this study provides insights into the internal characteristics
of a public bicycle system. The findings indicate that the usage of public bicycles is related to
both land use and the usage of bikes at nearby stations. Furthermore, the results show that
there are both high-demand and low-demand areas for public bicycles, with the average
service coverage meeting the intended purpose of "the first and last mile". These insights can
inform decision-making in the operation and maintenance of bike-sharing systems, improving
their efficiency and benefits for users.

The authors in [3] analyzed the Bike Share System (BSS) which is a sustainable and non-
motorized mode of transportation that is used for short to medium distance trips. The
"Location Problem" refers to finding the best locations for bike stations to maximize demand
coverage. Public-Private Partnerships are responsible for implementing BSS, but optimizing
location is important to maximize benefits due to budget constraints. They have used PIS and
NIS to determine the most and least important bike stations, respectively, and decision makers
can improve network efficiency by identifying new bike station locations or relocating the
least important ones. The methodology which is proposed in this paper combines the Maximal
Covering Location Problem (MCLP) and bike station importance to achieve the maximum
BSS demand coverage within a specified service distance, which is a pre-defined threshold
around a bike station. The three criteria used to measure the centrality of a BSS network are
Betweenness Centrality (BC), Closeness Centrality (CC), and Degree Centrality (DC). The
methodology applies Analytical Hierarchy Process (AHP) and Technique for Order
Preference by Similarity to Ideal Object (TOPSIS) to determine the weights of each criterion
and to rank the alternatives. The dataset used in this study is a historical data of one million
bike trips in the fourth quarter of 2018, obtained from the CBSS network website. The
network only covers Washington D.C and contains 99 bike stations. The paper presents a case
study to demonstrate the application of the new methodology and to compare its results with
existing methods.
This paper [4] focuses on analyzing the urban public transportation networks or systems
(PTN) of Hungarian five cities. Although unweighted networks were frequently the focus of
earlier research, this analysis novelty is that it takes into account directed and weighted links,
where the weights correspond to the carrying capabilities of the buses, trams, and trolleybuses
during morning rush hour. Global network characteristics like diameter, average path length,
eccentricity distribution, degree distribution, community structure, and, node centrality
measures like degree centrality, local average connectivity, closeness centrality, betweenness
centrality, page-rank centrality, extended centrality measures for weighted networks, were
calculated in both weighted and unweighted case. A highly detailed picture of the variations
in public transportation organization that may be caused by historical and geographical factors
was obtained by comparing the data obtained for the various cities. Additionally, by
contrasting the findings in the weighted and unweighted examples, the centralities were
compared, and overcrowded stops and underutilized stops with spare capacity were found.
The most vulnerable stations and routes in the network that highlight certain organizational
irregularities in the transportation system could also be found. They come to the conclusion
that the key structural components and organizing principles of the public transportation
systems can be revealed using the complex network approach and graph theoretic
measurements that were taken into account.

The authors of paper [5] argued that the structure of a city is largely dependent on the travel
patterns of its people. To support this argument, they have analyzed GPS-enabled taxi data
collected in Shanghai and China using complex network science methods to explore the
structure of intra-city flows. More attention is given to patterns of trips of shorter lengths
which are often preferred since they are affordable for majority residents in daily lives. The
final dataset contained attributes such as taxi ID, pick-up time, pick-up point coordinates,
drop-off tie, drop-off point coordinates and trip length. This data was converted into a graph
network with sources and destinations as nodes and trips between them as links. Community
detection from the network was done using the Infomap algorithm, which can handle
weighted and directed networks and perform quickly. To reveal the underlying structure, they
have first detected clustered sub-networks of short distance trips to find a basic structure and
slowly added trips of longer distances. After this, finally a two-level hierarchical nature of the
city structure was revealed. The density, node strength, betweenness centrality, closeness
centrality, normalized betweenness centrality and normalized closeness centrality are
calculated to analyze spatial interactions represented by taxi trips and to draw conclusions.

The identification of hierarchical structure and explanations of the underlying mechanisms

can contribute towards transportation planning of urban cities. The sub-regional borders
formed during clustering helped in identifying areas of local centers [5]. Therefore, policies
can be made to improve accessibility to these local centers, or to reduce the total amount of
travel. With cities becoming more complex, it is important that necessary policies are
implemented to improve the travel in the city.
The paper [6] presents a complex network analysis of passenger travel routes in rail and bus
transport public systems, in the island of Singapore. The network from this dataset contains
more than 4130 nodes. The characteristics of this data is compared with previously studied
transport systems like the Indian railway, Boston subway, Chinese railway and the worldwide
airport network. The study also highlights the importance of both topological and dynamical
properties of the networks. The degree, strength, clustering, assortativity and the eigenvector
centrality of this network are also calculated and reported. From this, it is noted that the
dynamical properties of the network varied differently from its topological properties. From a
topological view, the network is similar to a highly connected ER random graph, which tells
that almost all nodes are traveled from one to another. This indicates that the graph is almost
fully connected, with high clustering. But with dynamical analysis, a more complex system is
revealed, where some regions have higher traffic when compared to others. From the
eigenvector analysis, it was concluded that the traffic differs depending on the day of the
week, thus indicating the importance of temporal effects. But the study has a main
disadvantage, which is that it analyzes only a week’s data. When this analysis is performed on
data of larger scale, more inferences such as the differences in travel patterns among weeks or
years can be made. Although the study was done only on transportation data, it can also be
used on other classes of networks.

In this study [7], the network topology and nodal centrality of specific Chinese cities in the air
transport network of China(ATNC) are examined using a complex network technique. Degree
distribution, average path length, and clustering coefficient are metrics characterizing network
structure as a whole. The dataset came from the CAAC, or Civil Aviation Administration of
China (2009). This paper assesses the network structure measures like Degree distribution,
Average path length, Clustering coefficient. It calculates Centrality measures like Degree
centrality, Closeness centrality, Betweenness centrality and takes in account Correlation
measures like Degree correlation, Clustering-degree correlation. The degree, closeness, and
betweenness metrics for each city describe the advantages of a node's position as being
directly linked to others, accessible to others, and the mediator between others, respectively.
The results show that the ATNC exhibits some small-world (SW) network traits with an
average path length of 2.23 and a clustering coefficient of 0.69, and has a cumulative degree
distribution described by an exponential function. All three centrality indices have strong
correlations with socioeconomic city measures like population, gross regional domestic
product, and air passenger traffic (GRDP). The overall network structure and centrality of
each city in the ATNC has been examined in this study using network analysis indices, and
the geographical patterns of the cities are analyzed in connection to economic and geographic
aspects. This confirms that centrality captures an essential component of location advantage
in the ATNC and has significant effects on the spatial distribution of economic activities.
Proposed Work

1. Network Centrality Measure: Complex networks by definition are highly heterogeneous

structures. This often results in some parts of the network being more information-rich
than others. For example, in social networks, some individuals might have a large number
of connections and can spread information faster than others. Hence, within the context of
social network analysis, the nodes that represent them are considered more
important(central). Within the context of transportation networks, an urban region where
the influx of people is higher than in the other areas can be considered central. However,
the centrality of a given network may change over time as a result of the growth and
evolution of the underlying system. Thus, the definition of centrality is not absolute but
rather depends on the specific context and the purpose of the abstraction. Taking this
relativity into account, several centrality measures have been proposed that focus on
different types of relationships between the nodes.

1.1. Degree Centrality : “Degree” of a node refers to the number of nodes that a given
node is connected to. Degree centrality is the simplest measure of node connectivity.
Sometimes it’s useful to look at in-degree (number of inbound links) and out-degree
(number of outbound links) as distinct measures. Within the context of city bikes, this
refers to the number of bike stations that users have traveled to from the target station.
This could potentially help us in identifying whether any important place or hub is
nearby a bike station.

1.2. Betweenness Centrality: Betweenness centrality measures the number of times a

node lies on the shortest path between other nodes. Betweenness is useful for
analyzing communication dynamics. A high betweenness count could indicate
someone holds authority over disparate clusters in a network, or just that they are on
the periphery of both clusters. It helps us in finding important traffic points while
moving from one area to another area.

1.3. Closeness Centrality: Closeness centrality is a way of detecting nodes that are able
to distribute flows efficiently through the network. The closeness centrality is
calculated as the normalized average of all of its geodesic distances. Here, in our
module, a node with high closeness centrality will indicate stations that can
potentially act as versatile intermediary stations within the network. It helps us in
giving necessary inputs for network optimization.

1.4. Eigenvector Centrality: Eigen Centrality measures a node’s influence based on the
number of links it has to other nodes in the network. Eigen Centrality then goes a step
further by also taking into account how well connected a node is, and how many links
their connections have, and so on through the network. Eigenvector centrality in
individual transportation networks allows highlighting not only important singular
nodes but also geographically important areas within the city. This is possible
 because it acknowledges not only the importance of individual station but also the
importance of stations adjacent to them.

1.5. PageRank: PageRank is a variant of EigenCentrality, also assigning nodes a score

based on their connections, and their connections’ connections. The difference is that
PageRank also takes link direction and weight into account – so links can only pass
influence in one direction, and pass different amounts of influence.

2. Community Detection: Communities in networks refer to groups of nodes that are

densely connected internally. Community detection is often a crucial process for
understanding the structure of complex networks. In the case of the city bike network,
community detection can help to better understand bike usage patterns and determine
adequate pricing models and also about expanding new bike stations in those have dense
communities. This can be done by applying various methods like agglomerative and
divisive methods using various algorithms like Louvain Community Detection, Combo
Community Detection, etc.

Results and Discussion

The project overall analyses the complex bike network based on complex network theory and
using various visualizations. Centrality measures such as degree, betweenness, eigenvector,
and closeness will be analyzed to understand the characteristics of various nodes and their
importance in order to make various inferences. The usage of stations and their internal
relationship can also be understood through the analysis of strength and community structure
which is implemented using various community detection algorithms.

References

[1] Li, Ze-Tao, Wei-Peng Nie, Shi-Min Cai, Zhi-Dan Zhao, and Tao Zhou. "Exploring
the topological characteristics of urban trip networks based on taxi trajectory data."
Physica A: Statistical Mechanics and its Applications 609 (2023): 128391.
[2] Yao, Yi, Yifang Zhang, Lixin Tian, Nianxing Zhou, Zhilin Li, and Minggang Wang.
"Analysis of network structure of urban bike-sharing system: A case study based on
real-time data of a public bicycle system." Sustainability 11, no. 19 (2019): 5425.
[3] Salih-Elamin, R., & Al-Deek, H. (2021). A new method for determining optimal
locations of bike stations to maximize coverage in a bike share system network.
Canadian Journal of Civil Engineering, 48(5), 540-553.
[4] A. Háznagy, I. Fi, A. London and T. Nemeth, "Complex network analysis of public
transportation networks: A comprehensive study," 2015 International Conference on
Models and Technologies for Intelligent Transportation Systems (MT-ITS),
Budapest, Hungary, 2015, pp. 371-378, doi: 10.1109/MTITS.2015.7223282.
[5] Liu, Xi, et al. "Revealing travel patterns and city structure with taxi trip data."
Journal of transport Geography 43 (2015): 78-90.
[6] Soh, Harold, et al. "Weighted complex network analysis of travel routes on the
Singapore public transportation system." Physica A: Statistical Mechanics and its
Applications 389.24 (2010): 5852-5863.
[7] Exploring the network structure and nodal centrality of China’s air transport network:
A complex network approach. (2010, September 25). In Exploring the network
structure and nodal centrality of China’s air transport network: A complex network
approach - ScienceDirect. https://doi.org/10.1016/j.jtrangeo.2010.08.012