Community detection in financial hypergraphs

Abstract

The detection of communities is fundamental for elucidating the structural and functional organization of complex systems. Although community detection has been extensively investigated in pairwise financial networks, higher-order interactions among financial entities remain comparatively underexplored. In this study, we analyze community structures in financial hypergraphs by employing a homogeneity matrix that yields a weighted pairwise representation of higher-order interactions. Using daily closing price time series from the Bombay Stock Exchange, we construct triadic interaction networks and apply modularity maximization to the corresponding homogeneity graph to derive optimal community partitions. Our results indicate that communities arising from higher-order interactions exhibit pronounced temporal variability and differ substantially from both conventional sector-based classifications and pairwise interaction communities. Moreover, technology-sector firms tend to occupy central positions in such communities. The absence of persistent functional communities over time underscores the implications of the adaptive market hypothesis and reflects the intrinsically dynamical nature of financial markets. Copyright © 2026. Published by Elsevier Ltd.