Efficient Truss Computation for Large Hypergraphs

Abstract

Cohesive subgraph mining has been applied in many areas, including social networks, cooperation networks, and biological networks. The k-truss of a graph is the maximal subgraph in which each edge is contained in at least k triangles. Existing k-truss models are defined solely in pairwise graphs and are hence unsuitable for hypergraphs. In this paper, we propose a novel problem, named (k, α, β) -truss computation in hypergraphs. We then propose two hypergraph conversions. The first converts a hypergraph into a pairwise graph, while the second converts it into a projected graph. We further propose two algorithms for computing (k, α, β) -truss in hypergraphs based on these two types of conversions. Experiments show that our (k, α, β) -truss model is effective and our algorithms are efficient in large hypergraphs.