Analyzing Community-aware Centrality Measures Using The Linear Threshold Model (2202.00514v1)

Abstract: Targeting influential nodes in complex networks allows fastening or hindering rumors, epidemics, and electric blackouts. Since communities are prevalent in real-world networks, community-aware centrality measures exploit this information to target influential nodes. Researches show that they compare favorably with classical measures that are agnostic about the community structure. Although the diffusion process is of prime importance, previous studies consider mainly the famous Susceptible-Infected-Recovered (SIR) epidemic propagation model. This work investigates the consistency of previous analyses using the popular Linear Threshold (LT) propagation model, which characterizes many spreading processes in our real life. We perform a comparative analysis of seven influential community-aware centrality measures on thirteen real-world networks. Overall, results show that Community-based Mediator, Comm Centrality, and Modularity Vitality outperform the other measures. Moreover, Community-based Mediator is more effective on a tight budget (i.e., a small fraction of initially activated nodes), while Comm Centrality and Modularity Vitality perform better with a medium to a high fraction of initially activated nodes.