关键词: 层次结构, 重叠结构, 最大团, 社区发现

Abstract: The research shows that many real networks have hierarchical structure and overlapping structure. Traditional hierarchical clustering algorithms usually use the node as the object to expand and form the hierarchical tree graph to get the network hierarchy. This approach has two problems, one is the stability of the algorithm, mainly reflect in the selection of initial node, different initial nodes will lead to the final results of different algorithms in a few cases, even if the algorithm does not depend on the initial node, but the complexity of the algorithm will change, second is unable to find overlapping structure. To solve the above two problems, this paper proposes a hierarchical overlapping community detection algorithm based on maximum clique. The algorithm uses maximal cliques as the extended object and adopt maximal clique expansion strategy to generate the hierarchical tree diagram. Finally, the overlapping modularity function is used to prune the hierarchical tree and get the result of community division. The experimental results on real network and LFR artificial neural network show that the proposed algorithm can effectively mine the hierarchical structure and overlapping structure.

Key words: hierarchical structure, overlapping structure, maximum clique, community detection