ComSim : A bipartite community detection algorithm using cycle and node's similarity

Raphael Tackx (1) , Fabien Tarissan , Jean-Loup Guillaume (2)

1 ComplexNetworks
2 L3I - Laboratoire Informatique, Image et Interaction - EA 2118

This study proposes ComSim, a new algorithm to detect communities in bipartite networks. This approach generates a partition of nodes by relying on similarity between the nodes in terms of links towards ⊥ nodes. In order to show the relevance of this approach, we implemented and tested the algorithm on 2 small datasets equipped with a ground-truth partition of the nodes. It turns out that, compared to 3 baseline algorithms used in the context of bipartite graph, ComSim proposes the best communities. In addition, we tested the algorithm on a large scale network. Results show that ComSim has good performances, close in time to Louvain. Besides, a qualitative investigation of the communities detected by ComSim reveals that it proposes more balanced communities.

Mots clés: bipartite graph; social network; Community Detection

Domaines: Réseaux sociaux et d'information [cs.SI]

Raphael Tackx, Fabien Tarissan, Jean-Loup Guillaume. ComSim : A bipartite community detection algorithm using cycle and node's similarity. Complex Networks 2017, Nov 2017, Lyon, France. pp.278-289, ⟨10.1007/978-3-319-72150-7_23⟩. ⟨hal-01657093⟩