Topological Measures for Community Detection Assessment

These scripts implement several measures allowing to compare two community structures, i.e. two partitions of the node set of a given graph. They are based on popular measures defined in the field of cluster analysis, namely:

  • Purity (also known under many other names in the literature, such as percent correct, accuracy, etc.)
  • Rand index and its adjusted version.
  • Normalized mutual information.

The variants implemented here account for the network structure, an essential aspect of community structure which is otherwise completely ignored in standard measures.

  • URL: https://github.com/CompNet/TopoMeasures
  • Release date: 2012–2016
  • Related publications:
    • Vincent Labatut. « Generalized Measures for the Evaluation of Community Detection
      Methods ». In : International Journal of Social Network Mining (IJSNM) 2(1):44-63
      (2015). DOI: 10.1504/IJSNM.2015.069776. ⟨hal-00802923(cite this publication if you use this software)
    • Vincent Labatut. « Une nouvelle mesure pour l’évaluation des méthodes de détection
      de communautés ». In : 3ème Conférence sur les modèles et l’analyse de réseaux : approches mathématiques et informatiques (MARAMI). Villetaneuse, FR, 2012, 12p. MARAMIhal-00743888