An Enhanced Community Detection Method Based on Neighborhood Similarity

Detection of community structure in a social network is important for understanding the structure and dynamics of the network. Yet, most community detection algorithms do not take attributes of nodes and connections into consideration, or only make use of connections´ attributes information, not fully capturing the richness of the information contained in the data. Thus, by exploring the information of nodes and connections together in a social network, we propose an enhanced community detection method, the core of which is an agglomerative hierarchical clustering algorithm. The clustering algorithm bases on an improved evaluation algorithm of neighborhood similarity. Besides, the proposed method can provide several candidate partitions obtained according to the qualitative function modularity for users to choose. Finally, a real collaborative network of scientists is constructed with the data from DBLP and experiments on the network show that our method performs well.