26 JAN 2022 (WED) 16:05-16:35

Forecasting spatio-temporal spread of COVID-19 under different Interventions by using agent-based modeling with mobility data

Mr TANG Ka Chung, Ken PhD Student, Department of Geography, HKU


Sweeping the world since December 2019, the global pandemic of Corona Virus Disease 2019 (COVID-19) has been taking a severe toll on human lives and socio-economic activities globally. Governments across the globe have resorted to various non-pharmaceutical interventions (NPI), such as social distance, wearing a mask, and lockdown, and pharmaceutical intervention (PI), such as vaccination; however, the pandemic has not been contained nor controlled, as of January 2022. Therefore, identifying and designing effective interventions are crucial for managing the pandemic while requiring sophisticated intervention planning. Thus, an evidence-based forecast, which takes into consideration of the COVID-19’s epidemiological attributes and peoples’ response to COVID-19, to estimate and evaluate the effectiveness of various interventions is essential for intervention design. Many studies forecasted the trend of infected and deceased populations and assessed the effectiveness of NPIs by using mathematical modeling or statistical approaches at the national or city level. However, they are deficient in micro-scale prediction because of insufficient data for data training or parameterization. Furthermore, the evolving pandemic further hinders accurate forecasting by using historical data. Agent-Based Modeling (ABM) is a powerful technique for simulating spatial-temporal dynamics of population, pandemics, and various interventions at different scales. Although many studies adopted ABM to forecast the spread of COVID-19, most of them only focus on calculating infection/fatality probability based on demographic information. Alternatively, the “SCEIQDR” model (Susceptible, Close Contacted, Exposed, Infectious, Quarantine, Deceased and Recovered model), an extension of the “SEIR” (Susceptible, Exposed, Infectious, Recovered) epidemiological model, has been developed based on an extended disease ecology model to incorporate the factors of COVID-19 using publicly available data including Transportation Characteristic Survey 2011 (TCS 2011), Census and Geospatial Data from the government. Hence, population, individual behavior, and habitat could be comprehensively considered in ABM by using this epidemiological model. This study will increase the accuracy of mobility patterns by classifying mobility pattern clusters and generating an appropriate trip pattern using the latest Census data (i.e., Census data 2016) with machine learning techniques. Finally, this study will explore future public health implications based on the forecasting result in this study to manage the risk for future pandemic prevention and control.