关键词: 社交网络, 层次社区, 拓扑势, 峰谷结构

Abstract: Social networks often demonstrate hierarchical community structures. Most traditional agglomerative hierarchical community detection methods are often inefficient, and the generated dendrograms are complex. Motivated by solving these problems, the paper proposes a novel hierarchical community detection method on topology potential. The method employs the natural peak-valley structure of topology potential field to reveal the hierarchical relationship among communities in social network. The proposed method firstly identifies the local maximal potential nodes and detects the initial local community structure of the network on the basis of these nodes. After that, the initial local communities are iteratively merged according to the distance among the maximal potential nodes until all the communities are merged into one community. Experimental results on both synthetic and real-world networks show that the proposed method can discover the hierarchical community structure efficiency, and the generated dendrogram is simple and intuitive.

Key words: social network, hierarchical community, topology potential, peak-valley structure