An improved community detection algorithm based on modified local expansion method

Abstract

Community detection is an important research issue in complex network mining. In this paper, firstly, we define central nodes, called Extended Local Max-Degree (ELMD) nodes in a complex network. All the central nodes are used for the community expanding. We also prove that ELMD method is more precise and dispersed than local max-degree method in the real datasets. Secondly, we propose an improved local expansion method to expand community from the seeds (ELMD nodes), and this process is named as Extended Local Community Expansion with Modified R method (ELCEMR). ELCEMR is an unsupervised learning method, and does not need any priori-knowledge. Finally, the validations against the real-world datasets show that the proposed method performs better than other algorithms for community detection.