Finding communities in bicycle sharing system

Abstract

The bicycle sharing system (BSS) provides a more sustainable transport paradigm in big cities. The recorded cycling trajectories can be used to detect human movement patterns. Community detection methods have been used to study BSS from a complex network perspective. However, the previous used modularity-based methods not only ignored the interdependencies of bicycle flows in the system, but also suffered from the problem of resolution limit. The in-depth analysis of detection results is also lacked. In this paper, we propose an interactive visual analytics system to detect the cycling communities of bicycle sharing system. Different kinds of community detection algorithms are adopted for finding station clusters; multiple inter-linked views are designed to visualize properties of the detected substructures from different perspectives. The real bicycle sharing dataset in Hangzhou is used for analysis, which demonstrates the effectiveness of our method. By using the system, analyzers can compare the cluster results generated by different algorithms, investigate the reason of the partition results based on different metrics, and find the relationship among human activity communities and the city subregional structures. This study provides insights into using bicycle sharing data to reveal human travel pattern and BSS usage pattern, which potentially aids in developing urban planning policies.