FCA for Common Interest Communities discovering

Abstract

Major scientific and industrial issues related to social networks have led many researchers to focus on the longstanding problem of automatic communities extraction. The vast majority of the proposed methods tend to make a partition of the entities from the initial graph of observed relationships. The semantics of these relationships is rarely considered. The lack of information about the elements that connect or repel individuals course, is the main cause. In this article, we focus our interest on the detection of Common Interest Communities (CIC) in social networks. In this respect, we introduce a new approach called SOC Miner for communities detection based on the use of both Formal Concept Analysis (F CA) techniques and data from social networks. Carried out experiments over real-world data sets emphasize the relevance of our proposal and open many issues.

Multidimensional community discovering in heterogeneous social networks

“…To overcome this shortage, many researches have focused on detecting communities in homogeneous networks based on semantic information and nodes attributes 15 . There are four categories of approaches for semantic community detection in social networks, 16 depending on their input: a clustering‐based approach, 12-14,17 a model‐based approach, 15-19 a label propagation based approach, 20,21 and a graph‐based approach 7,22-25 …”

“…As a consequence, the new community detection challenge consists in combining the structural information with users' attributes (ie, the semantic information). For this reason, many research proposed approaches for monodimensional community detection based on formal concept analysis ( FCA ) techniques, 7,8 where nodes of a lattice represents a community and an edge represents the subsumption relation.…”