Coarse-grained diffusion distance for community structure detection in complex networks
Coarse-grained diffusion distance for community structure detection in complex networks
复制标题
DOI：
10.1088/1742-5468/2010/12/p12030
复制
发表时间：
2010-12
期刊：
影响因子：
--
通讯作者：
Jian Liu;Tingzhan Liu
中科院分区：
文献类型：
--
作者：
Jian Liu;Tingzhan Liu
代表真实系统的复杂网络最相关的特征之一是社区结构。本文通过特征映射将网络中节点间的扩散距离测度推广到具有数据参数化的粗粒度网络上的广义形式。粗粒度网络中元节点的接近性概念反映了在扩散过程中社区连通性方面划分的内在几何形状。然后，在此度量下，通过聚集分层聚类技术将节点分组成社区，并使用模块化函数选择生成的树图的最佳分区。该算法能够以较高的效率和精度识别社区结构。在没有任何关于社区结构的先验知识的情况下，可以自动确定适当数量的社区。在几个人工网络和实际网络上的计算结果证实了该算法的有效性。
One of the most relevant features of complex networks representing real systems is the community structure. In this paper, we extend the measure of diffusion distance between nodes in a network to a generalized form on the coarse-grained network with data parameterization via eigenmaps. This notion of proximity of meta-nodes in the coarse-grained networks reflects the intrinsic geometry of the partition in terms of connectivity of the communities in a diffusion process. Nodes are then grouped into communities through an agglomerative hierarchical clustering technique under this measure and the modularity function is used to select the best partition of the resulting dendrogram. The present algorithm can identify the community structure with a high degree of efficiency and accuracy. An appropriate number of communities can be automatically determined without any prior knowledge about the community structure. The computational results on several artificial and real-world networks confirm the capability of the algorithm.