计算机应用与软件2018，Vol.35Issue(2)：256-260,315,6.DOI:10.3969/j.issn.1000-386x.2018.02.046
一种基于图金字塔的聚类算法
A CLUSTERING ALGORITHM BASED ON GRAPH PYRAMID
摘要
Abstract
The traditional k-means method and hierarchical clustering algorithm produce incorrect clustering results when there is an outlier or cluster overlap in a data set.Inspired by wavelet multiresolution analysis,this paper proposed a clustering method based on graph pyramid.Firstly, the number of classes in dataset K was input, and the minimum spanning tree was constructed for the data points.Then,the priority was calculated according to the degrees of the nodes and the length of the shortest neighbor.Next, the minimum spanning tree was traversed from high to low by priority. Finally,the clustering results formed by the merged nodes were output.The experimental results on the synthesized and true data sets showed that compared with k-means and hierarchical clustering method, the clustering results of this method were not affected by the overlap between outliers and clusters and had high stability.关键词
聚类/金字塔/多分辨率/最小生成树/层次聚类Key words
Clustering/Pyramid/Multiresolution/Minimum spanning tree/Hierarchical clustering分类
信息技术与安全科学引用本文复制引用
吕晓波,马燕,张相芬,李顺宝,张玉萍..一种基于图金字塔的聚类算法[J].计算机应用与软件,2018,35(2):256-260,315,6.基金项目
国家自然科学基金项目(61373004) （61373004）
校创新团队项目(A700115001005) （A700115001005）
校基金项目(Sk201220). （Sk201220）
