基于局部相似性的K-means谱聚类算法

Algorithm of K-means Spectral Clustering Based on Local Similarity

定义科学的局部相似性指数是基于局部相似性社团发现算法的关键,根据共有邻居信息定义的局部相似性指数对直接相连节点对的相似性数值存在低估倾向,本研究将节点对的关联信息加入到局部相似性指数的定义中,结合K-means谱聚类算法对网络节点进行聚类.本研究定义的局部相似性指数克服了传统局部相似性指数的缺点,且保持了原有的计算复杂性.在计算机生成网络和实际网络上运行,并和经典算法做了比较,实验证明,所提算法能够较为有效、准确地检测网络的社团结构.

The scientific definition of the local similarity index is essential for the algorithm of community detection based on local similarity. The local similarity indexes based on common neighbors underestimate the similarity value of neighbor nodes, The correla- tion information of node pairs is involved in the definition of local similarity index, network nodes are clustered by this similarity measure combining with K-means spectral clustering. The similarity index proposed by the paper overcomes the shortcomings of tradi- tional local similarity index, and maintains the original computational complexity . The proposed method is tested on both computer- generated and real-world networks, and is compared with the typical algorithms in community detection. Experimental results verify and confirm the feasibility and validity of the proposed method.