Hierarchical Agglomeration Community Detection Algorithm via Community Similarity Measures

Community detection is an important method for analyzing the community structure of real-world networks. Most of the hierarchical agglomeration community detection algorithms are the variations of NM algorithm. In this contribution, we present a new hierarchical agglomeration community detection algorithm, called Community Merging via Community Similarity Measures (CMCSM). The proposed algorithm encompasses three components. It first repeatedly joins communities by using single-node community measure and combination rule. Then it adjusts a few nodes by SHARC which is an advanced label propagation algorithm. Finally, it merges communities by using community similarity measure. Four of most important features of CMCSM are that (1) it requires only a single parameter which is the number of community count, (2) it can prevent single-node communities and monster communities from being created, (3) it is well suited for a wide range of networks and (4) its computation is not expensive. The algorithm CMCSM is demonstrated with real-world and artificial networks, the experiment shows that CMCSM has a more efficient and accurate result of community detection compared with some hierarchical algorithms recently proposed.