基于划分的聚类分析算法的改进  

The improvement of division-based clustering analysis algorithm

作　　者：范多锋[1] 徐俊刚[1] 

机构地区：[1]中国科学院研究生院,北京100097

出　　处：《微型机与应用》2012年第18期54-56,59,共4页Microcomputer & Its Applications

摘　　要：对传统的K-平均算法作了简单的介绍和讨论,提出了一种具有单纯型法思想的K-中心点轮换法。分别对比了K-均值算法与K-中心点轮换算法的时间复杂度,针对K-中心点轮换算法的时间复杂度提出了一种基于抽样原理的改进算法,并对K-中心点轮换算法聚类数目的选择进行了各种改进方法的探索。同时,基于主流的weka开源数据挖掘工具实现了改进算法。实验结果表明了算法的有效性。This paper makes a brief presentation and discussion of the traditional K-means algorithm, and then proposes a simplex method thought the K-center rotation algorithm, K-Mediods. This article compares the time complexity of K-means algorithm with K-center point of rotation algorithms. Finally, this paper proposes an improved algorithm based on sampling theory for the time complexity of the K-center rotation algorithm, and the choice of the number of the K-center rotation algorithm clustering the exploration of the various improvements, and then explores the various improvements for the choice of the clustering number of the K-center rotation algorithm. Meanwhile, this paper achieves the improved algorithm based on weka. which is a mainstream open source data mining tools. The experimental results show the effectiveness of the algorithm.

关 键 词：K-均值 K-中心点轮换 抽样 聚类数目 WEKA 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]