An Iterated Heuristic Community Detection Algorithm for Social Networks Based on Centrality and Similarity Measures

Résumé

Community detection is a crucial analytical approach to comprehend the structure and functionality of intricate networks, which are frequently conceptualized as graphical representations. This problem is extremely difficult and has not yet been satisfactorily solved. It is believed that neighbors may strongly surround a community's central or leader node, and that the centers of two communities may be far apart. Furthermore, it is postulated that the similarity between nodes within the same community is greater than that observed between nodes belonging to different communities. Thus, it is evident that local and global structural information is important in community detection. This paper presents an iterative heuristic algorithm for community detection in social networks using centrality and similarity measures called HCCS, which consists of four main steps: parameter initialization, leader nodes' selection, communities' formation, and post-processing. The main contributions are as follows: first, both local and global information about the network is considered, and then a heuristic formula for measuring the similarity between two nodes is redefined. The leader nodes are identified based on two criteria: degree and distance. During the community formation phase, the degree of similarity between the nodes is quantified using the heuristic formula. The nodes are then assigned to the same community once they reach maximum similarity. The postprocessing phase entails the integration of two communities when the modularity increment is positive, reaching its maximum when the two communities are merged. Experiments on real networks demonstrate the effectiveness of the proposed approach.

Citer ce document

Accès au document

Texte intégral en lecture en ligne, réservé aux abonnés SPHAERO et aux membres de l'institution. Se connecter

Voir l'article sur le site de la revue

Auteur(s)

Statistiques

Consultations : 1

Téléchargements : 0