基于最优特征向量的谱二分社团检测方法

发布时间：2019-06-24 23:02

【摘要】：针对传统谱二分社团检测算法一般只使用某一特定的特征向量对网络进行划分,并不能保证能够得到最佳的社团结构这一缺陷,提出了一种使用最优特征向量的谱二分社团检测方法。该方法利用网络/子网络转移矩阵的特征向量持续将网络分裂为若干个子网络,分裂过程并不固定使用单一的、特定的特征向量,每次分裂使用的是能使得模块度增量最大的一个特征向量。此外,为了充分利用网络的拓扑信息,还利用网络中每条边所关联的两个顶点拥有的共同邻居的信息,将原始网络转换为带权的网络,并基于此带权网络的转移矩阵,使用最优特征向量持续将其划分为若干个子网络,得到其社团结构。为了验证这两种方法的有效性,在7个实际网络上进行了实验。实验结果证实,该方法能够有效地从网络中提取高质量的社团结构。
[Abstract]:In view of the fact that the traditional spectral dichotomous community detection algorithm generally only uses a specific eigenvector to divide the network, which can not guarantee the optimal community structure, a spectral dichotomous community detection method using the optimal eigenvector is proposed. In this method, the network is continuously divided into several subnetworks by using the eigenvector of the network / subnetwork transfer matrix. The split process does not use a single, specific eigenvector. Each split uses a eigenvector that can maximize the modularity increment. In addition, in order to make full use of the topological information of the network, the original network is transformed into a weighted network by using the information of the common neighbors owned by the two vertices associated with each edge of the network, and based on the transfer matrix of the weighted network, the optimal eigenvector is used to continuously divide it into several subnetworks, and its community structure is obtained. In order to verify the effectiveness of the two methods, experiments are carried out on seven practical networks. The experimental results show that this method can effectively extract high quality community structure from the network.
【作者单位】：兰州大学信息科学与工程学院;甘肃省资源环境科学数据工程技术研究中心;
【基金】：2015年甘肃省资源环境科学数据工程技术研究中心开放基金~~
【分类号】：O157.5

【相似文献】