Spatiotemporal analysis of bike mobility chain: A new perspective on mobility pattern discovery in urban bike-sharing system

Bike-sharing data have been a valuable source for urban transport research. While most studies focus on Origin-Destination (OD) data of users' bike trips in bike-sharing researches, little has been investigated on mobility patterns from the perspective of bikes. Bike mobility can not only reflect human mobility patterns but also provide practical insights for understanding the bike-sharing system and assisting bike management. In particular, continuous bike movement information can provide abundant spatiotemporal connection information for spatial context analysis in data without user ID information. This paper studies bike mobility by shifting the perspective from OD analysis to the new analysis primitive of bike mobility chain (BMC). First, bike-sharing OD trip data are reconstructed to obtain BMCs. Second, the reconstructed BMCs are feature-augmented, based on which the spatiotemporal characteristics of bike mobility are analyzed. Finally, the word embedding model and the clustering algorithm are applied to the BMCs, which embody rich spatiotemporal inter-station connectivity, to facilitate mining stations with similar bike mobility. The methodology is applied for the comparative study of bike-sharing activities on workdays and holidays using data collected from New York Citi-Bike. The experimental results manifest locality patterns of bike mobility in both mobility chain indicators and multi-scale station embedding analysis results, and reveal disparities in BMC statistic indicators and spatial distribution mode of stations with similar bike mobility. These results can provide a new empirical reference for urban bike management in different time periods. As a more general implication, this study broadens the perspective for OD-based data analysis and paves the path to leverage the thriving trajectory-based research for further investigations.