The Structure of Spatial Networks and Communities in Bicycle Sharing Systems

Abstract

Bicycle sharing systems exist in hundreds of cities around the world, with the aim of providing a form of public transport with the associated health and environmental benefits of cycling without the burden of private ownership and maintenance. Five cities have provided research data on the journeys (start and end time and location) taking place in their bicycle sharing system. In this paper, we employ visualization, descriptive statistics and spatial and network analysis tools to explore system usage in these cities, using techniques to investigate features specific to the unique geographies of each, and uncovering similarities between different systems. Journey displacement analysis demonstrates similar journey distances across the cities sampled, and the (out)strength rank curve for the top 50 stands in each city displays a similar scaling law for each. Community detection in the derived network can identify local pockets of use, and spatial network corrections provide the opportunity for insight above and beyond proximity/popularity correlations predicted by simple spatial interaction models.

bikedata

“…All other comparable sources of data (such public transit data, or mobile phone data) are either not publicly available, or have been released only at single distinct times for single distinct purposes. Many public hire bicycle systems in the U.S.A., along with Santander Cycles in London, U.K., issue ongoing releases of their usage data, providing a unique source of data for analysing, visualising, and understanding human movement and urban environments (Austwick 2013;Borgnat et al 2011;Padgham 2012). Such data provide an invaluable resource for urban planners, geographers, social and health scientists and policy makers, data visualisation specialists, and data-affine users of the systems themselves.…”

Enhancing Space-Aware Community Detection Using Degree Constrained Spatial Null Model

“…They use a modified version of the Louvain algorithm [5] to optimize their variant of modularity. Several articles, for instance [6,7], applied this approach on different case studies.…”

“…The deterrence function is first computed from these data, when needed, and the degree of nodes is used as proxy for the intrinsic importance of nodes, as it is often done in applications to collected datasets, for instance in [7,8]. This setting is more realistic, for applications to real world datasets.…”

Detecting Urban Dynamics with Taxi Trip Data for Evaluation and Optimizing of Spatial Planning:

“…Compared with modern western urban planning, which quantitatively analyzes and solves urban problems through building a vigorous index system (Alexander, 2000;Benevolo & Landry, 1967), spatial planning in China has been relatively more subjective and drawn from much attention to aesthetics. Traditional spatial planning has been trying to manage individual travel behavior by means of mere construction of physical environment.…”

“…As community detection can be considered a process for dividing complex networks into several sub networks, among which the internal connections are extraordinarily closed, many researchers have emphatically studied the formation process of communities. Gao et al (2013), Roth et al (2011), Liu, Y. et al (2014, Austwick et al (2013) and Liu, X. et al (2015) all focused on this topic, however, all of these researchers failed in evaluating the urban practices with revealed urban dynamics phenomena.…”