Pré-Publication, Document De Travail Année : 2025

Flexible Optimization of Structural Measures for Community Detection

Résumé

Community detection is a crucial task in many graph data analysis pipelines, with applications ranging from social networks to biology and transportation systems. This work introduces FO-SM-CD , a novel constraint programming framework for community detection. While existing approaches mainly rely on the modularity criteria to detect communities, FO-SM-CD allows for the combination of different structural measures such as conductance, density, and cut-ratio. It also supports user-defined constraints, providing unprecedented flexibility to tailor the detection process to specific needs. On top of an agglomerative algorithm to community detection strategy, this article makes two major contributions. First, it provides a constraint-based modeling of different structural measures that can be combined in the objective function, and second, it exploits a connectedness property to reduce the search space and speed up the identification of optimal partitions. Extensive experiments on controlled synthetic datasets confirm the diversity of graph partitions obtained when different structural measures are used. It also demonstrates the computational efficiency of the connectedness constraint combined with iterative local search heuristics compared to existing approaches.

Fichier principal
Vignette du fichier
_DKE_25__FO_SM_CD (2).pdf (471.27 Ko) Télécharger le fichier
Origine Fichiers produits par l'(les) auteur(s)
Licence

Dates et versions

hal-05299688 , version 1 (06-10-2025)

Licence

Identifiants

  • HAL Id : hal-05299688 , version 1

Citer

Arnold Hien, Cécile Bothorel, Laurent Brisson, Nicolas Jullien, Patrick Meyer, et al.. Flexible Optimization of Structural Measures for Community Detection. 2025. ⟨hal-05299688⟩
1808 Consultations
233 Téléchargements

Partager

  • More