北京航空航天大学学报 ›› 2009, Vol. 35 ›› Issue (12): 1499-1502.

• 论文 • 上一篇    下一篇

基于Squeezer算法的大规模矩阵聚类分析

李 岩, 王惠文, 叶 明, 刘 丹   

  1. 北京航空航天大学 经济管理学院, 北京 100191
  • 收稿日期:2009-02-15 出版日期:2009-12-31 发布日期:2010-09-14
  • 作者简介:李 岩(1979-),男,河南正阳人,博士生,buaaly@126.com.
  • 基金资助:

    国家自然科学基金资助项目(70771004,70521001,70531010)

Modifiable Squeezer cluster algorithm used in large-scale matrix

Li Yan, Wang Huiwen, Ye Ming, Liu Dan   

  1. School of Economics and Management, Beijing University of Aeronautics and Astronautics, Beijing 100191, China
  • Received:2009-02-15 Online:2009-12-31 Published:2010-09-14

摘要: 对Squeezer算法进行分析研究,在定义2个矩阵之间距离的基础上,提出了一种改进的Squeezer算法,用于对维数相同的大规模矩阵进行聚类分析.改进的算法在设定距离阈值的基础上,对类别的半径设定阈值来控制分类精度,给出具体的算法步骤来实现针对大量矩阵的聚类分析.对聚类后所得矩阵集合,给出集合质心和半径的定义,来描述矩阵集合的特性.所提算法能使聚类结果避免受到链条效应的影响而使类不断扩容,从而导致聚类精度下降的问题.仿真实验分析验证了所提算法具有良好的聚类效果和适用性.

Abstract: To solve the clustering method to the large-scale matrixes in the same dimension, the modifiable Squeezer cluster algorithm was proposed, based on the analysis of Squeezer cluster algorithm and the definition of the distance between the matrixes. The modifiable algorithm set a distance threshold, put forward a threshold of radius to control the accuracy of classification, and gave the detailed algorithm steps to realize cluster analysis for a large number of matrices. When the matrix cluster set was obtained, the modifiable algorithm provided the definition of center and radius to describe the properties of the matrix set. The proposed method could control the accuracy of classification in order to prevent chain effect in the course of clustering. The simulation experiment was addressed to validate the rationality and effectiveness of the modifiable algorithm.

中图分类号: 


版权所有 © 《北京航空航天大学学报》编辑部
通讯地址:北京市海淀区学院路37号 北京航空航天大学学报编辑部 邮编:100191 E-mail:jbuaa@buaa.edu.cn
本系统由北京玛格泰克科技发展有限公司设计开发