Addressing COVID-induced changes in spatiotemporal travel mobility and community structure utilizing trip data: An innovative graph-based deep learning approach

The COVID-19 pandemic has resulted in significant disruptions in mobility patterns, leading to changes in user travel behavior. Understanding users’ travel demand, travel behaviors, and changes in the structure of the travel network becomes the basis for governments and operators to provide improved service quality. Public transportation in a city provides essential mobility, accessibility, and connectivity for residents. The burgeoning shared mobility sector utilizes the Internet to establish a management platform that leverages digital technology to offer convenient travel services. Bike sharing presents a new transport mode for short-distance trips strengthening connectivity with public travel modes such as buses and subways, while online taxi services take on long-distance trips within cities. This paper proposes a network-based deep learning method to address the COVID-induced changes in spatiotemporal travel mobility and community structure detection, which integrates graph learning and optimization in an end-to-end training approach. The approach involves constructing a dynamic travel network and adopting complex network theory to develop metrics that uncover the changes in user mobility patterns and explore the correlation between different travel modes. Our results show that the pandemic reduces overall trip volume and network structure changes, suggesting that productive and residential activities have partially recovered but remain far from pre-pandemic levels, especially for taxi and subway trips. These findings provide valuable insights for transportation planners and policymakers to explore strategies that promote more sustainable and resilient mobility patterns in the post-pandemic era.