SP-MHDS: Realistic Population Contact Network with Consideration of Spatial Locality

The foundation of improving the accuracy of infectious disease transmission simulations lies in realistic population contact networks. Yet, acquiring large-scale, high-resolution contact data with spatial details is challenging, and existing synthetic network models lack real-world characteristics such as multi-layered structure, heterogeneity, dynamics, and small-world nature. They also miss spatial constraints on contact formation, failing to reflect human interaction patterns accurately. Addressing this, we propose the SP-MHDS (SynthPops-Multilayer-Heterogeneous-Dynamic-Small World) method for constructing contact networks in disease simulation. This method enhances SynthPops-generated social networks with a community layer, utilizing social networks as prior knowledge to extract potential spatial distribution of nodes through modularity Q calculation and node-merging strategies. It establishes stochastic, clustered, and dynamic edges within community layers under spatial constraints. Visualization and experimental results affirm that SP-MHDS has robust spatial clustering and small-world properties. Utilizing Covasim, a widely recognized model for simulating COVID-19 spread, we conducted a simulation of infection trends among the population in King County, Seattle, from June to December 2020. A comparative analysis with actual epidemiological figures affirmed the superior verisimilitude of the SP-MHDS contact network compared to the combination of SynthPops and other conventional networks, validating its heightened appropriateness for expansive, spatially sensitive epidemic simulations.