| 聚类分析的一种改进方法 |
| |
| 引用本文： | 朱惠倩. 聚类分析的一种改进方法[J]. 湖南文理学院学报(自然科学版), 2005, 17(3): 7-9,16 |
| |
| 作者姓名： | 朱惠倩 |
| |
| 作者单位： | 华东交通大学,基础科学学院,江西,南昌,330013 |
| |
| 摘 要： | 为了削减由于指标变量不同量纲、不同权重及其相关性对聚类效果造成的影响,把权重引入马氏距离计算公式, 定义了加权马氏距离,并证明了其在理论上的有效性. 最后,采用加权马氏距离对一个具体算例进行聚类分析,得到了较为理想的、更符合实际的分类结果,表明了该定义在应用上的有效性.
|
| 关 键 词： | 聚类分析 相关性 权重 马氏距离 |
| 文章编号： | 1672-6146（2005）03-0007-03 |
| 收稿时间： | 2005-04-20 |
| 修稿时间： | 2005-04-20 |
An Improvement on Cluster Analysis |
| |
| ZHU Hui-qian. An Improvement on Cluster Analysis[J]. Journal of Hunan University of Arts and Science:Natural Science Edition, 2005, 17(3): 7-9,16 |
| |
| Authors: | ZHU Hui-qian |
| |
| Abstract: | To diminish the influence on clustering effect resulting from different measurements, different weights and correlation of indicator variables. weights are brought in Mahalanobis distance in this paper. Then weighted Mahalanobis distance is defined and proved valid in theory. At last, a clustering analysis is then conducted for a specific case based on weighted Mahalanobis distance, which generates more reasonable and more genuine classified result, indicates the validity of weighted Mahalanobis distance in practice. |
| |
| Keywords: | cluster analysis correlation weight Mahalanobis distance |
| 本文献已被 CNKI 维普 万方数据 等数据库收录！ |
|
