Generating synthetic population for simulating the spatiotemporal dynamics of epidemics

Agent-based models are effective for simulating epidemic spread, but their accuracy depends on realistic population data. When comprehensive data is lacking, synthetic populations are crucial. This study introduces a novel method for generating synthetic populations tailored for infectious disease transmission simulations. It integrates household structures from microsamples and uses a heuristic combinatorial optimizer to recalibrate these structures, creating a spatially-explicit synthetic population of over 17 million agents for Shenzhen, China. The method effectively replicates statistical structural relationships, aligning with demographic benchmarks. The study also shows that variations in population synthesizers significantly alter epidemic projections, impacting peak incidence rate and onset.

Kemin Zhu, Ling Yin, Kang Liu, Junli Liu, Yepeng Shi, Xuan Li, Hongyang Zou, Huibin Du