摘要
图聚类是基于各种标准如结点标号、边标号、公共子图等条件将图数据集实例划分不同类集群,这将对结构化图空间及增强对图数据的理解有着重要作用。针对此问题提出基于结构化的图聚类算法。与目前有关的算法相比,该算法不产生新图或原图分解成零碎子图,也不依赖计算最大共同子图的相关操作。实验结果表明,这种方法在现实分子图数据集上对结构聚类可行、有效。
The graph clustering is to partition the instances in graph datasets into different clusters based on various criteria such as node labels, edge labels and common subgraph. This will play important role to structuring the graph space and to comprehending the graph data. Aiming at this issue we present the structuring-based graph clustering method. In contrast to existing related approaches, this method does not generate new graph or decompose the original graph into fragmentary subgraphs, and does not rely on the correlated operation of computing the maximum common subgraph （MCS） as well. Experimental results show that such method is feasible and effective in clustering the structures on real molecular graph dataset.
出处
《计算机应用与软件》
CSCD
北大核心
2014年第6期18-20,58,共4页
Computer Applications and Software
基金
国家高新技术研究发展计划项目(2012AA112312)
教育部高等学校博士学科点专项科研基金项目(20110161120006)
湖南警察学院自然科学基金项目(2011YB02)
关键词
集群
结构聚类
图数据集
频繁子图挖掘
Clusters Structural clustering Graph datasets Frequent subgraph mining
