Top influencers can be identified universally by combining classical centralities

Abstract: Information flow, opinion, and epidemics spread over structured networks. When using individual node centrality indicators to predict which nodes will be among the top influencers or spreaders in a large network, no single centrality has consistently good ranking power. We show that statistical classifiers using two or more centralities as input are instead consistently predictive over many diverse, static real-world topologies. Certain pairs of centralities cooperate particularly well in statistically drawing the boundary between the top spreaders and the rest: local centralities measuring the size of a node's neighbourhood benefit from the addition of a global centrality such as the eigenvector centrality, closeness, or the core number. This is, intuitively, because a local centrality may rank highly some nodes which are located in dense, but peripheral regions of the network---a situation in which an additional global centrality indicator can help by prioritising nodes located more centrally. The nodes selected as superspreaders will usually jointly maximise the values of both centralities. As a result of the interplay between centrality indicators, training classifiers with seven classical indicators leads to a nearly maximum average precision function (0.995) across the networks in this study.