IEEE Circuits and Systems Magazine - Q4 2019 - 62

Extracting useful information from huge and distributed datasets remains
a major challenge. In public transport network analysis, the size
of datasets, typically consisting of several thousand nodes, is relatively
midget and the time for data mining is also comparatively manageable.
output files which record the footprints of every
vehicle during the simulation time at a sampling
rate of 1 sec. By evaluating the maximum speed
for every road segment and the geographic distance between the stops, the end-to-end travel
delay can be calculated using (25).
Our results showed the dependency of the vehicular speed along a road segment upon the node
weight, as discussed in Section IV-J. Specifically,
we observed that the higher the node weight, the
lower the maximum speed attained by the vehicles
on the road segment, especially during rush hours.
The speed was observed to be further affected
when the distance between the stops is reduced.
Our simulation results have been verified using real-world data provided by the Kowloon Motor Bus
Co., one of the major transport operators in Hong
Kong [13]. Fig. 8(b) shows the dependency of the
maximum speed attained along a road segment
(Vmax ) for a normalized node weight ^w i_normh. Empirical data are in good agreement with our simulation results. We may conclude that with increased
node weight (demand) and reduced geographical
distance between the stops, the attainable speed
by vehicles along a road segment is reduced significantly. In practice, when the bus stops are located
closer to each other to offer better service, traffic speed will be compromised, and more aggressive reduction of distance between the stops may
even lead to a state of traffic congestion. Our node
weight model can be adopted to facilitate a better
route planning and stop deployment to maintain
optimal traffic performance.
VI. Conclusion and Future Work
In a data driven world, the availability of real-world datasets
and high-end tools for handling huge datasets has greatly
facilitated the research of complex system and data analysis. Extracting useful information from huge and distributed datasets remains a major challenge. In public transport
network (PTN) analysis, the size of datasets, typically consisting of several thousand nodes, is relatively midget and
the time for data mining is also comparatively manageable.
Despite the successful attempts in applying concepts from
network science to PTN analysis, serious study of PTN
from a network science perspective is still relatively rare.
62

IEEE CIRCUITS AND SYSTEMS MAGAZINE

In this paper, we aimed at bringing together some of the recent developments in the application of network theory to
PTN analysis. In particular, useful contributions have been
made by various researchers in the use of L-space representation in comparison to P-, B- and C-space representations, since the L-space graph structure mimics the actual
real-world infrastructure of a PTN. A directed and weighted
network structure is best suited for the study of bus transport structures, whereas an undirected and weighted network structure is more suited for metro transport studies,
and the main reason for considering the graph type is the
level of overlapping among inbound and outbound routes.
We have found that the notion of supernodes offers practical and more insightful perspective to understanding the
actual network behavior, which is difficult to be captured
by conventional graph representations. Furthermore, adding static weights to nodes and edges has been found to
be effective in capturing the significance of nodes and links
in PTNs. It is worth noting that merely representing the
PTN structure as a graph and analyzing various network
parameters may not lead to practically useful conclusions
because the purpose of the public transport systems is to
meet travel needs of the community being served, which
requires the consideration of more practical network parameters. Also, considering the spatial embedding of PTNs
alongside with the topological analysis conveys more insightful information without which quantifying the network
might yield rudimentary results.
Topological analysis of PTNs have been performed
using various local metrics (e.g., degree, clustering, betweenness centrality, closeness centrality), global metrics
(e.g., degree distribution, scale-free property, average path
length, small-world property), and pairwise properties
(e.g., assortativity and communities). The study of various
local, global, and pairwise properties has provided intriguing information about the topological behavior of public
transport networks. Such study has provided a great
source of information for researchers in the applied fields,
for example, in designing of transfer algorithm, optimization of public transport routes, prediction and regulation
of road congestion, network planning, transit operation,
etc. However, while PTN analysis generates information
like the existence of hierarchical structure, core-periphery
structure, and the absence of scaling in a PTN, such information does not find immediate practical relevance to the
PTN operators or government agencies. Thus, more work
FOURTH QUARTER 2019
IEEE Circuits and Systems Magazine - Q4 2019

Table of Contents for the Digital Edition of IEEE Circuits and Systems Magazine - Q4 2019
