基于候选聚类的K调和均值算法(KHM-CC)  

K Harmonic Means Algorithm Based on Candidate Clustering Method(KHM-CC)

作　　者：陈改霞 王建平[2] 

机构地区：[1]鹤壁汽车工程职业学院,河南鹤壁458030 [2]河南科技学院信息工程学院,河南新乡453003

出　　处：《火力与指挥控制》2016年第1期48-52,共5页Fire Control & Command Control

基　　金：国家自然科学基金(31371525);河南省教育厅科学技术研究重点基金(14A520067);河南省教育厅人文社会科学研究基金(2014-GH-245);河南省信息技术教育研究重点基金(ITE12037);2014年度新乡市科技发展计划基金(14GY23);2014年度河南科技学院教育教学改革研究重点基金资助项目(2014PUZD08)

摘　　要：聚类分析是常见的数据分析技术。基于KHM的聚类分析是当前研究的热点。提出了基于候选聚类的KHM算法(KHM-CC),详细阐述了算法的设计过程。采用UCI的小样本(iris)数据集和大样本(Bag of Words)数据集对比了KHM-CC算法和禁忌搜索KHM算法(KHM-TS)和变邻域搜索KHM算法(KHM-VNS)的性能。实验结果表明,KHM-CC算法在处理iris数据集小样本数据集时,其性能和KHM-VNS算法基本接近,而优于KHM-TS算法。但是在处理Bag of Words大样本数据集时,性能优于KHM-VNS和KHM-TS算法,其聚类计算耗时明显缩短,证实KHM-CC算法在高维度数据集的处理上更具优势。Clustering analysis is a common data analysis technology. The Clustering analysis Based on KHM clustering is a hotspot of current research. In this paper, a KHM algorithm based on candidate clustering technology （KHM-CC） is proposed. The designing process of the algorithm is expournded, and uses small sample and large sample of UCI to compare KHM-CC, KHM-TS and KHM-VNS. The experimental results show that the performance of KHM-CC and KHM-VNS algorithm is close for treating the iris data set,and the result is better than that of KHM-TS. But in dealing with the large sample data sets （Bag of Words）,the performance is better than that of KHM-VNS and KHM-TS,the computing time of clustering is significantly shortened. It is confirmed that KHM-CC has advantage on the processing of high dimensional data set.

关 键 词：聚类分析 数据分析 候选聚类KHM算法(KHM-CC) 禁忌搜索KHM算法(KHM书) 变邻域搜索KHM算法(KHM-VNS) 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]