Hypergraph Community Detection Based on Higher-Order Topology and Information Flow

Jing Xiao, Zhi-Wei Ma, Jing Cao, Xiaoke XuIEEE Transactions on Computational Social Systems 2026 · DOI

Abstract

Hypergraph community detection has gained significant attention due to reflecting higher-order mesoscale structures and facilitating functional analysis. However, previous methods mainly rely on a single and biased hypergraph property, i.e., static hyperedge structures or dynamic information flow, but largely overlook their intrinsic cooccurrence and potential conflicts in real-world situations. To address this issue, we first model hypergraph community detection as a biobjective optimization problem by considering the two inherent and ubiquitous community properties, and design two conflicting objectives, i.e., improved average hypergraph conductance and codelength. Second, to bridge higher-order interactions with evolutionary search, we introduce a topology-preserving weighted graph reduction strategy that transforms complex hyperedges into manageable super-dyadic weights. Finally, in order to identify more comprehensive, diverse, and explanatory hypergraph partitions, a multiobjective hypergraph community detection framework based on high-order topology and information flow (MOHCD-HI) is proposed. Specifically, MOHCD-HI follows the classical multiobjective evolutionary optimization framework to approximate a set of pareto-optimal partitions. Moreover, a novel neighbor-hyperedge-based mutation (NH-mutation) incorporates the hyperedge neighborhood topologies of nodes to expand the mutation space for more flexible and sufficient variation, facilitating the generation of higher-quality and more diverse offsprings. Experimental results demonstrate that MOHCD-HI outperforms state-of-the-art algorithms in terms of higher-order topology and information flow indicators, thereby effectively reflecting intrinsic properties while offering a macroscopic perspective for decision-making.
BibTeX
@article{xiao2026hypergraph,
  author = {Jing Xiao and Zhi-Wei Ma and Jing Cao and Xiaoke Xu},
  title = {{Hypergraph Community Detection Based on Higher-Order Topology and Information Flow}},
  journal = {IEEE Transactions on Computational Social Systems},
  year = {2026},
  doi = {10.1109/tcss.2026.3655885},
}

Loading related content...

Figure