Understanding urban dynamics using graph-based analysis of public bike-sharing schemes

Urbanisation has contributed to more dense and diverse travel flows and interaction between people and urban spaces. In recent years, mobility systems have incorporated more Information technology and sensors, where the movement of people can be sensed and captured at high frequency. The increased availability of such data enables people and their interactions with the physical environment to be analysed and understood within the context of complex urban systems. It also has the potential to be used for evidence-based urban planning and management. However, problems such as congestion and air pollution continue to happen as results of high car-dependency in many cities. To overcome the challenges, sustainable modes of transportation such as cycling is (re)emerging through bike-sharing schemes as an increasingly common response. Analysing flow interactions is key to promoting and developing sustainable and integrated transportation systems in cities.
This study has aimed to better understand and model dynamics within sustainable urban transport systems through spatiotemporal and novel graph-based analysis. Large quantities of travel flows in different and newly emerged (e.g.dockless) bike-sharing schemes are investigated with high spatial and temporal granularity at the individual level. Weekly, daily and hourly graph structures are modelled to uncover and quantify how travel behaviours are impacted by the built environment, by changes in transportation systems, and as a result of events that disrupt the normal patterns of movement. The revealed mobility patterns and graph structural changes indicate underlying urban resilience, where individuals can rapidly adopt bike-sharing to fill transit gaps that emerged. Furthermore, this thesis, for the first time, quantifies the utility of temporal graph indices for enhancing machine learning models to making predictions on dynamics and demand in urban traffic systems. This research has great potential to help improve transport management and planning in urban areas.