Measuring node spreading power by expected cluster degree

Abstract

Traditional metrics of node influence such as degree or betweenness identify highly influential nodes, but are rarely usefully accurate in quantifying the spreading power of nodes which are not. Such nodes are the vast majority of the network, and the most likely entry points for novel influences, be they pandemic disease or new ideas. Several recent works have suggested metrics based on path counting. The current work proposes instead using the expected number of infected-susceptible edges, and shows that this measure predicts spreading power in discrete time, continuous time, and competitive spreading processes simulated on large random networks and on real world networks. Applied to the Ugandan road network, it predicts that Ebola is unlikely to pose a pandemic threat.