Boolean factor based community extraction from directed networks with the non reciprocal link relationship

Abstract

The community extraction from the networks has gained more attention this last decade due to the available data provided by the online social media. This task consists in extracting the homogeneous groups from the network modelled by a graph. The graph models the interaction between entities of the network through the edges. Most of the existing approaches of the community extraction have been designed for the non-directed graph or considered that the relationship between entities is symmetric or reciprocal. In most of the real world application like food web or hierarchical relationship between employees, it is not the case. In this paper, we propose a boolean factor based approach for community detection in directed networks. The main advantage of the boolean factor, based on formal concepts is that, it keeps the relationship between the two sets of related entities. The semantic relationship (non reciprocal) is taken into account during the candidate community extraction process by splitting concept into two parts. The final communities is obtained after refinement of these candidates. We have experimented this approach on some collected directed networks available on internet and the results show the effectiveness of this approach.