More about HKUST
Managing Urban Crises via the Lens of Human Mobility Data: Measurement and Intervention
The Hong Kong University of Science and Technology Department of Computer Science and Engineering PhD Thesis Defence Title: "Managing Urban Crises via the Lens of Human Mobility Data: Measurement and Intervention" By Miss Lin CHEN Abstract: Cities are complex sociophysical systems marked by heterogeneity, interconnectivity, and dynamicity, where people, infrastructures, and opportunities are unevenly distributed, mutually influenced, and constantly evolving. These features pose unique challenges in tackling urban crises such as pandemics and natural disasters, which require a comprehensive understanding of what people experience and how they react in complex urban environments. Human mobility data, collected from widely-adopted GPS-equipped mobile devices, provide a powerful study lens with its large-scale, fine- grained, and real-time nature. In this thesis, I develop data-driven methodologies that harness human mobility data for sensing, understanding, and intervening in urban crises. First, to sense vulnerable populations in urban crises, I construct VulnerabilityMap, an open data framework that integrates large-scale mobility patterns with demographic, environmental, and behavioral indicators. Drawing on a taxonomy of shocks and stresses inspired by Maslow's hierarchy of needs, VulnerabilityMap produces fine-grained measures of community disadvantage and crisis impact. Machine learning models trained on this data demonstrate strong predictive performance and reveal key social factors that trap populations in vulnerable circumstances. Second, to understand the progression of urban crises, I conduct a thorough data-driven analysis across different phases of the COVID-19 pandemic at scale. I identify two dominant mobility response patterns: daily life-centered venues experience smaller reductions and quicker recovery, while work-centered venues face larger reductions that persist. These changes exert deeper impacts on the underlying social fabric, significantly correlating with income and racial segregation. Clustering analysis further uncovers geographical heterogeneity in mobility responses, linked to partisan inclinations. Third, to understand the role of mobility network structure in shaping social resilience during urban crises, I construct city-wide mobility networks based on visitation preference dependencies. Network centrality emerges as a robust predictor of social resilience, explaining over 80% more variance in both segregation and mobility changes than traditional metrics. Moreover, core and peripheral places identified from the network do not distinguish each other by their static geographic distribution, but instead differ dramatically in how, when, and from where people visit them—a result with direct implications for targeting crisis-response interventions. Fourth, to intervene effectively and equitably, I develop a demographic- and mobility‑aware epidemic simulation model and an optimization framework for resource allocation, demonstrated through COVID-19 vaccine distribution. Simulation results show that prioritizing disadvantaged communities improves both overall social utility and equity even amid vaccine hesitancy, and that multi-objective optimization can reconcile conflicts between different equity dimensions. Finally, I conclude this thesis with future research directions and challenges related to data-driven urban crisis management. Date: Friday, 25 April 2025 Time: 2:45pm - 4:45pm Venue: Room 5562 Lifts 27/28 Chairman: Prof. Ajay JONEJA (IDEA) Committee Members: Dr. Tristan BRAUD (Supervisor) Prof. Pan HUI (Supervisor, EMIA) Prof. James KWOK Prof. Qiong LUO Dr. Gareth TYSON (ECE) Prof. Flora SALIM (UNSW)