摘要
针对大多数现有的聚类有效性函数都是针对于数值型数据提出的,无法有效地评价和分析类属型数据的问题,提出了一种新的聚类有效性函数—修正划分模糊度;通过结合模糊划分熵和划分模糊度测度,所提出的聚类有效性函数既可以评价数值型数据分类结果,也可以评价类属型数据的分类性能。
The cluster validity is an important topic of cluster analysis, which is often converted into the determination of the optimal cluster number. Most of the available cluster validity functions are limited for the analysis of numeric data set and ineffective for the categorical data set. For this purpose, a new cluster validity function is presented, namely the modified partition fuzzy degree. By combining the partition entropy and the partition fuzzy degree, the new cluster validity function can be applied to any data set with numeric attributes or categorical attributes. The experimental results illustrate the effectiveness of the proposed cluster validity function.
出处
《系统工程与电子技术》
EI
CSCD
北大核心
2005年第4期723-726,共4页
Systems Engineering and Electronics
基金
国家自然科学基金 (60 2 0 2 0 0 4)资助课题
关键词
聚类分析
聚类有效性函数
数值特征
类属特征
cluster analysis
cluster validity function
numeric attributes
categorical attributes
