Policy and Newly Confirmed Cases Universally Shape the Human Mobility during COVID-19

Abstract

Understanding how human mobility pattern changes during the COVID-19 is of great importance in controlling the transmission of the pandemic. This pattern seems unpredictable due to the complex social contexts, individual behaviors, and limited data. We analyze the human mobility data of over 10 million smart devices in three major cities in China from January 2020 to March 2021. We find that the human mobility across multi-waves of epidemics presents a surprisingly similar pattern in these three cities, despite their significant gaps in geographic environments and epidemic intensities. In particular, we reveal that the COVID-19 policies and statistics (i.e., confirmed cases) dominate the human mobility during the pandemic. Thus, we propose a universal conditional generative adversarial network based framework to estimate human mobility, integrating COVID-19 statistics and policies via a gating fusion module. Extensive numerical experiments demonstrate that our model is generalizable for estimating human mobility dynamics accurately across three cities with multi-waves of COVID-19. Beyond, our model also allows policymakers to better evaluate the potential influences of various policies on human mobility and mitigate the unprecedented and disruptive pandemic.