论文部分内容阅读
研究复杂网络中的社区结构,有助于发现网络结构和功能的关系,进而理解复杂网络的组成规律、预测复杂网络的行为.文章基于支持向量机的思想和LM(Louvain Method)非重叠社区发现算法,提出一种采用邻居投票机制的LM-NV(Louvain Method with Neighbor Voting)重叠社区发现方法,基本思想是保留非重叠社区的部分结构,采用一种基于局部信息的邻居投票机制仅对社区边界节点的社区隶属情况进行判别.LM-NV算法易于扩展到大规模复杂网络,同时不存在对社区个数的初始化问题.在基准测试网络和真实网络上的实验结果表明LM-NV算法不仅具有良好的时间效率,而且在社区发现准确度上优于其它代表性算法.
Studying the community structure in complex networks helps to discover the relationship between network structure and function, and then comprehends the composition rules of complex networks and predicts the behavior of complex networks.Based on the idea of support vector machines and the Louvain Method (LM) non-overlapping community discovery Algorithm, this paper proposes a Louvain Method with Neighbor Voting (LM-NV) overlapping community discovery method based on neighbor voting mechanism. The basic idea is to preserve the partial structure of non-overlapping communities. A neighbor voting mechanism based on local information is used only for community boundaries Node community membership.LM-NV algorithm is easily extended to large-scale complex network, and there is no initialization problem on the number of communities.Experimental results on benchmark network and real network show that LM-NV algorithm not only has good Time efficiency, but also outperforms other representative algorithms in community discovery accuracy.