Influence Study on Hyper-graphs

Multilateral relations between entities lose their semantics when represented as simple graphs. Instead hypergraphs can naturally represent the said relations, which are common in social tagging systems. An important issue is the effect of the structural properties of a hypergraph on influence propagation. In the current work, an empirical study is undertaken to compare the effect of degree, k-shell and eigenvector centrality under the SIS, and SIR models of infection. The results on the MovieLens, Delicious and LastFM social networks indicate that k-shell centrality is a more accurate predictor of the influence of a node than degree centrality, and that eigenvector centrality is closely correlated with k-shell centrality.