Resilience of urban complex sociotechnical systems
How can we model the complex recovery dynamics of coupled sociotechnical systems after shocks? I propose and test system dynamics modeling and big data assimilation approaches -- find out more in my papers in PNAS, Sustainable Cities and Society, and Environment and Planning B.
Post-disaster population displacement and recovery analysis
Quantifying the dynamic flows of population displacement after disasters is crucial for response and recovery. How can use mobile phone data, which has its limitations and biases? See our papers in Royal Society Interface, PLoS ONE, & review paper in Computers, Environment, and Urban Systems.
Geospatial AI for urban analytics and cross-city learning
Deep learning on spatio-temporal data enables us to learn high-dimensional representations of places, trajectories, and regions. How can we transfer insights across cities to predict the black swan events? See our papers in Nature Machine Intelligence, KDD2019, KDD2020, SIGSPATIAL2019.
COVID-19 mobility analysis and outbreak prediction
Were the non-compulsory stay-at-home orders in Japan effective in restricting mobility patterns in the early stages of the pandemic? We use mobility data and web search data provided by Yahoo Japan to analyze and predict outbreaks in Tokyo. For more, see our papers in Scientific Reports, CEUS, and KDD2022.
Measuring social capital via diversity of social connections in cities
How did the quality of urban physical encounters change during the pandemic and beyond? How do social connectivity and segregation affect community recovery after disasters? See our papers in Nature Communications, Transport. Research D, Applied Network Science, and Natural Hazards Review.
Human mobility simulations during shocks via reinforcement learning
Simulating city-scale population movements during disasters is crucial for urban management. How can we learn irregular behavior from past disasters to predict future events? See our papers using inverse reinforcement learning in Transportation Research Part C, SIGSPATIAL2020, & SIGSPATIAL2018.
Inferring economic resilience of businesses via human mobility
How can we infer the causal impact of businesses due to disaster events? We proposed and tested a Bayesian structural time series approach in Puerto Rico, in our paper in EPJ Data Science. I've also wrote a blog piece on implementation of this model using Python and stan. [More work coming soon!]
World Bank Consultancy:
Data for Development
I am also passionate in translating research into practice. I have been working with the Global Facility for Disaster Reduction and Recovery (GFDRR) at the World Bank to bring data-driven methods into project operations. We developed Mobilkit, an open-source toolkit for mobility data analytics with MindEarth.