摘要
社区结构发现研究可揭示复杂网络中隐藏中观结构,为进一步开展网络的形成和演化研究应用提供依据,如可为智能推荐、舆情控制、电力和交通网络调度等方面提供决策支持数据。针对复杂网络社区结构挖掘中社区数量难以确定的问题,提出一种基于密度自适应聚类数的社区发现谱方法。引入谱图分析中比较成熟的谱聚类特征向量分析方法,基于局部节点密度构图,结合网络图的边介数值构造相似矩阵,规范化后进行谱聚类,求得最大特征维度k值,k值即为社区个数。最后采用k-means方法对特征向量空间进行聚类,使得复杂网络社区得以呈现。在人工UCI和真实数据集(southern women data)上的实验表明,与现有谱聚类社区发现算法相比,该方法能自动确定社区个数,能得到划分精度更高的社区。
The study of community structure discovery can reveal the hidden meso-structure in complex network,and provide a basis for further research and application of network formation and evolution,such as providing decision support data for intelligent recommendation,public opinion control,power and traffic network scheduling.In view of the difficulty in determining the number of communities in complex network community structure mining,we propose a community detection spectral clustering method based on density adaptive generation cluster number.Based on the local node density composition,the similarity matrix is constructed with the boundary value of the network graph.After normalization,the spectral clustering is carried out to obtain the maximum characteristic dimension k value which is the number of community.Finally,the k-means method is used to cluster the eigenvector space,which makes the complex network community presented.Experiments on artificial and real data sets (UCI and Southern Women Data) show that this method can automatically determine the number of communities and obtain communities with higher division accuracy compared with the existing spectral clustering algorithm.
作者
王学军
李有红
李炽平
WANG Xue-jun;LI You-hong;LI Chi-ping(Huali College Guangdong University of Technology,Guangzhou 511325,China)
出处
《计算机技术与发展》
2019年第5期81-85,共5页
Computer Technology and Development
基金
广东高校省级重点平台和重大科研项目(2015KQNCX221)
