Ego-centered community detection in directed and weighted networks

Community detection is one of the most studied topics in Social Network Analysis. Research in this realm has predominantly focus on finding out communities by considering the network as a whole. That is, all nodes are put in the same pool to define central metrics for finding out communities while ignoring the particularity of some nodes and their impact. Yet, if the position of some nodes matters when defining the metrics (i.e. node centric approach), the found communities may differ and can make more sens in real life situations. For instance, identifying the communities based on drug dealers and their interactions with others sounds better than finding communities while ignoring the individuals status. The purpose of this paper is to detect ego-centered community, which is defined as a community built from a particular node. Our solution is set to combine both link direction and weight, and therefore, differs from many existing solutions. Basically, we rely on a metric called a quality function that uses link properties to assess the cohesion of identified groups. Our method detect communities that reflect not only the structure but the reality regarding to the interaction nature in terms of intensity. We implement our solution and use "Les Miserables" dataset to demonstrate the effectiveness of our solution.