Community Detection in Social Network with Node Attributes Based on Formal Concept Analysis

As far as social networks are concerned, new applications appear to analyze them. Community detection is one of the most important issues. It allows to understand the structure of complex networks and to extract useful information from the detected communities. Users have usually a social interaction with their friends because of their common interests or their similar profiles. In this paper, attributed graphs are considered, where entities of the network are described using attributes with several modalities. Then, we propose an hybrid approach based on Formal Concept Analysis (FCA) for community detection in social network with node attributes. This method, called ACDC (Attributed Community Detection based on Concepts), combines the structure of the network and the attributes of the nodes. ACDC, semantically and statically, partitions an assigned graph into k densely connected communities, using maximal cliques, with homogeneous attribute values derived from FCA. Experimental results demonstrate the effectiveness of ACDC through comparison with the state-of-the-art graph clustering and methods. Our method provides also more meaningful communities than conventional methods that consider only relationship information.