从链接密度遍历序列中挖掘网络社团的层次结构

提出一种称为TRAVEL的网络聚类算法.它能够产生包含所有可能密度聚类的网络链接遍历序列,并从中自动发现网络的全局优化聚类.然后,遍历序列被转挟为连续子区间堆结构.在此基础上,提出一种聚类算法HCLU,可以无须用户干预地从连续子区间堆中自动发现网络的层次聚类边界.在真实网络以及计算机生成的仿真网络数据集上的实验结果表明...

Mining Hierarchical Community Structure Within Networks from Density-Connected Traveling Orders

This paper proposes a density-based network clustering algorithm, TRAVEL. The algorithm produces a traveling order containing clustering with various densities and finds the optimal clusters in it. The traveling order is subsequently transformed into a data structure of contiguous subinterval heap based on which a clustering algorithm, HCLU, is designed to find the hierarchical cluster boundaries of the network without any user interaction. Experimental results on real-world and computer-generated synthetic networks show that the clustering accuracy of the proposed algorithms is higher than the baseline methods. Furthermore, they are able to produce robust hierarchical communities in various networks with low redundancy in the presence of noise.