«上一篇/Previous Article|本期目录/Table of Contents|下一篇/Next Article»

《计算机技术与发展》[ISSN:1006-6977/CN:61-1281/TN]

卷:

26

期数:

2016年06期

页码:

65-68

栏目:

智能、算法、系统工程

出版日期:

2016-06-10

文章信息/Info

Title:

 Study of Clustering Validity Evaluation on Semi-supervised Clustering Algorithm with Feature Discrimination

文章编号:

1673-629X（2016)06-0065-04

作者:

 李龙龙[1][2][3]; 何东健[2]; 王美丽[4]

1. 陕西工业职业技术学院 信息工程学院；2. 西北农林科技大学 机械与电子工程学院；3. 英国诺丁汉大学 计算机学院；4.西北农林科技大学 信息工程学院

Author(s):

 LI Long-long[1][2][3]; HE Dong-jian[2]; WANG Mei-li[4]

关键词:

 ; 聚类有效性; 半监督聚类; 算法评估; 成对约束; 最佳聚类数

Keywords:

 ; clustering validity; semi-supervised clustering; algorithm evaluation; pairwise constraints; optimal clustering number

分类号:

TP182

文献标志码:

A

摘要:

 鉴于最佳聚类数在提高聚类算法性能并扩大其应用领域方面的重要性，为了有效解决聚类算法中最佳聚类数的确定问题，解决传统的聚类分析算法常常需要人为预先指定聚类数的缺点，文中提出一种新型模糊半监督加权聚类算法。首先使用该算法对实测数据进行聚类，获取聚类结果。随后采用4种模糊聚类有效性评价算法依次对不同聚类数下的聚类结果进行聚类分析，最终通过不同聚类评价结果的对比分析得到实验数据的最佳聚类数。自测数据集的相关实验结果表明，不同的聚类有效性评价算法具有不同的优缺点，选择合适的聚类评价算法能够有效地解决最佳聚类数的确定问题，并能够有效提高实测数据的聚类识别率。

Abstract:

 As the optimal clustering number has great importance in improving the performance of clustering algorithm and expanding the algorithm’s application area,in order to solve the problem of the determination of the optimal clustering number for clustering algorithms effectively and settle the problem that the traditional clustering algorithm often requires prespecified number of clustering,a novel semi-supervised fuzzy clustering algorithm with feature discrimination ( SFFD) is proposed. Firstly,it is used to obtain the clustering result of the measured data,and then four kinds of fuzzy clustering validity evaluation algorithm are adopted for clustering analysis under different clustering number. Finally,by the comparative analysis of various validity evaluation algorithm with experimental data the optimal cluste-ring number was obtained. The experiment based on self-test datasets shows that various clustering validity evaluation algorithm has both the advantages and disadvantages,making a good choice for the clustering validity evaluation algorithm can effectively handle the problem of the determination of the optimal clustering number and enhance the recognition rate effectively for the measured data.

常用功能

更新日期/Last Update: 2016-09-20