Community detection in complex network based on an improved random algorithm using local and global network information

Abstract

These days, there are many types of complex networks that understanding the topology and functions of them allows us to derive valuable information from these networks. In this regards, community detection is an important research area that divides network graph into several subsets of network nodes called communities. The nodes included in each community are densely communicated to each other and are sparsely communicated to those nodes outside of this community. In this paper, we propose a novel community detection method that uses local and global network information resulting in low complexity and high accuracy. The our proposal is performed based on our identified architecture composed of four components including Pre-Processing, Primary Communities Composing, Communities Merging and Best Community Structure Selecting components. In the first component, we identify and store similarity measures and assign appropriate weights to network nodes and links based on local network information. Then, the second component considers similarity measures to compose a primary community structure based on a random algorithm improved by nodes’ weights. In the third component, we merge primary communities to achieve different community structures. Finally, the fourth component selects the best community structure considering evaluation functions calculated based on local and global network information. We evaluate our proposal based on different cases of real and artificial networks. Results show that our proposal can detect communities similar to real communities and has acceptable and efficient evaluation functions in all networks with any size and type compared with other proposals.