《计算机应用研究》|Application Research of Computers

基于划分和层次的混合动态聚类算法

Hybrid dynamic clustering algorithm based on partition and hierarchical clustering

免费全文下载 (已被下载 次)  
获取PDF全文
作者 郝洪星,朱玉全,陈耿,李米娜
机构 1.江苏大学 计算机科学与通信工程学院,江苏 镇江 212013;2.南京审计学院 信息科学学院,南京 211815
统计 摘要被查看 次,已被下载
文章编号 1001-3695(2011)01-0051-03
DOI 10.3969/j.issn.1001-3695.2011.01.012
摘要 针对划分聚类对初始值较为敏感以及层次聚类时间复杂度高等缺陷,提出了一种基于划分和层次的混合动态聚类算法HDC-PH。该算法首先使用划分聚类快速生成一定数量的子簇,然后以整体相似度的聚类质量评价标准来动态改变聚类数目,同时给出了聚类过程中孤立点的剔除方法。实验结果表明,HDC-PH算法的性能明显优于划分和层次算法,提高了聚类质量,并获得了更自然的聚类结果。
关键词 K-means;CURE;混合聚类;孤立点;整体相似度
基金项目 江苏省“青蓝工程”;江苏省六大人才高峰项目(07-E-025);江苏省高校自然科学重大基金研究项目(08KJA520001);国家中小企业创新基金资助项目(09C26213203797);国家自然科学基金资助项目(70971067)
本文URL http://www.arocmag.com/article/1001-3695(2011)01-0051-03.html
英文标题 Hybrid dynamic clustering algorithm based on partition and hierarchical clustering
作者英文名 HAO Hong-xing, ZHU Yu-quan, CHEN Geng, LI Mi-na
机构英文名 1. School of Computer Science & Telecommunications Engineering, Jiangsu University, Zhenjiang Jiangsu 212013, China; 2. School of Information Science, Nanjing Audit University, Nanjing 211815, China
英文摘要 For resolving the problem that partition clustering algorithms are sensitive to initial value and the time complexity of hierarchical clustering algorithms is high, this paper proposed a new hybrid dynamic clustering algorithm called HDC-PH, which was based on partition and hierarchical method.At first, HDC-PH partitioned the input data set into a number of subclusters, and then in a hierarchical manner, continuously merged the subclusters dynamically by using overall similarity to evaluate the cluster quality. At the same time, described a method to explain how to eliminate outlier during clustering. Experimental results show that HDC-PH is better than partition and hierarchical algorithms and get the better clustering results.
英文关键词 K-means; CURE; hybrid clustering; outlier; overall similarity
参考文献 查看稿件参考文献
 
收稿日期
修回日期
页码 51-53
中图分类号
文献标志码 A