Contagion backbone of temporal higher-order networks
TL;DR
It is shown, for the Susceptible-Infectious threshold process on temporal higher-order networks derived from human face-to-face interactions, that the contribution of each hyperlink can be quantified by a contagion backbone, whose dependency on the diffusion parameters is demonstrated and supported by theoretical analysis.
Abstract
Temporal higher-order networks, where each hyperlink involving a group of nodes is activated or deactivated over time, effectively represent social interactions. They serve as substrates for the spread of epidemics and information. However, the contribution of each hyperlink to a contagion process, namely, the average number of nodes that are infected via its activation, and the network properties of hyperlinks that influence this contribution, remain unexplored. Here we show, for the Susceptible-Infectious threshold process on temporal higher-order networks derived from human face-to-face interactions, that the contribution of each hyperlink can be quantified by a contagion backbone, whose dependency on the diffusion parameters is demonstrated and supported by theoretical analysis. We design centrality metrics of hyperlinks to estimate hyperlink rankings based on their contributions, revealing that local properties of hyperlinks can effectively identify high-contributing hyperlinks, and explain why different centrality metrics perform better under different process parameters. These insights are crucial for designing effective interventions that mitigate the spread of epidemics or misinformation.