Research On Community Detection Based On Modularity Maximization

With the rapid development of the social network, the research of complex networks has attracted much attention in many areas. As community structure is regarded as a common feature of the complex network, mining community structure becomes a research hotspot in the network analysis. A complex network is composed of a number of communities, and community detection is of great significance to understand the community structure. From the view of the calculation, community division can decompose the large scale task to reduce the computational complexity. While in the practical application, the recommendation in applications is based on community detection. And community detection facilitates people’s lives. In this paper, the research on the time complexity and the strength of community structure is based on the method of modularity maximization and the important attribute of community structure.Firstly, “merger going after label propagation” is utilized in this paper in order to reduce time complexity. Updating the whole network after communities merging every time can result in the high time complexity. CDMM-LPA(Community Detection based on Label Propagation Algorithm with Modularity Maximization) is proposed in this paper by combing label propagation with community structure. CDMM-LPA uses community structure because of its great significance and the guarantee in strong sense community.Secondly, CDPM(Community Detection based on Pairwise Merging of sub-communities) is proposed in this paper. It is an advanced algorithm based on community detection in social networks using merging sub-communities. CDPM uses similarity to decrease the number of sub-communities in primary stage. The executions of CDPM are decided by the check of community structure. So, there is no parameter in CDPM dominating the executions and the running time is cut down in the following.Finally, empirical analysis on real networks and artificial data networks verifies the validity of our approaches.