A Non-overlapping Community Detection Approach Based on $$\alpha $$-Structural Similarity

Community detection in social networks is a widely studied topic in Artificial Intelligence and graph analysis. It can be useful to discover hidden relations between users, the target audience in digital marketing, and the recommender system, amongst others. In this context, some of the existing proposals for finding communities in networks are agglomerative methods. These methods used similarities or link prediction between nodes to discover the communities in graphs. The different similarity metrics used in these proposals focused mainly on common neighbors between similar nodes. However, such definitions are missing in the sense that they do not take into account the connection between common neighbors. In this paper, we propose a new similarity measure, named $$\alpha $$ -Structural Similarity, that focuses not only on common neighbors of nodes but also on their connections. Afterwards, in the light of $$\alpha $$ -Structural Similarity, we extend the Hierarchical Clustering algorithm to identify disjoint communities in networks. Finally, we conduct extensive experiments on synthetic networks and various well-known real-world networks to confirm the efficiency of our approach.