现有大多数的网络聚类方法都只是针对无向网络,已有的有向网络聚类方法建立在传统聚类算法基础之上,存在着一定的局限性.针对上述问题,提出一种基于仿射传播的有向网络聚类算法,该算法首先采用Sim-Rank作为节点之间的相似度,并将计算得到的结果转换为适应于仿射传播算法的负值；然后将相似度矩阵作为输入,利用具有更好性能的仿射传播算法对有向网络进行聚类.实验结果表明,所提出算法的聚类性能优于其他几种具有代表性的有向网络聚类算法.%Most of the current clustering algorithms are designed only for undirected networks.The few existing algorithms for directed networks are designed based on traditional clustering methods which have some drawbacks.For these problems,this paper proposed a clustering method for directed networks using affinity propagation.Firstly,it calculated the SimRank similarities between the pairs of nodes in the given directed network.Then it converted the similarities into the equivalent negative value.Finally,it partitioned the networks into clusters using the famous and outstanding affinity propagation algorithm taking the calculated similarity matrix as input.The experimental results demonstrate that the performance of the proposed algorithm is superior to the state-of-art methods.
展开▼
