A local multiresolution algorithm for detecting communities of unbalanced structures

•We propose a fast local expansion algorithm for community detection named LSE.•It uses quality function that enables detecting communities independently of their sizes that can be of different densities.•The proposed algorithm can detect multiresolution community from a source vertex.•The experimental results verify that LSE can uncover rich information on networks.

In complex networks such as computer and information networks, social networks or biological networks a community structure is a common and important property. Community detection in complex networks has attracted a lot of attention in recent years. Community detection is the problem of finding closely related groups within a network. Modularity optimisation is a widely accepted method for community detection. It has been shown that the modularity optimisation has a resolution limit because it is unable to detect communities with sizes smaller than a certain number of vertices defined with network size. In this paper we propose a metric for describing community structures that enables community detection better than other metrics. We present a fast local expansion algorithm for community detection. The proposed algorithm provides online multiresolution community detection from a source vertex. Experimental results show that the proposed algorithm is efficient in both real-world and synthetic networks.