计算机工程与应用 ›› 2009, Vol. 45 ›› Issue (1): 156-158.DOI: 10.3778/j.issn.1002-8331.2009.01.049

• 数据库、信号与信息处理 • 上一篇    下一篇

基于网格相对密度的多密度聚类算法

程国庆,陈晓云   

  1. 福州大学 数学与计算机科学学院,福州 350002
  • 收稿日期:2007-12-25 修回日期:2008-03-03 出版日期:2009-01-01 发布日期:2009-01-01
  • 通讯作者: 程国庆

Clustering algorithm for multi-density based on grid relative density

CHENG Guo-qing,CHEN Xiao-yun   

  1. College of Mathematics and Computer Science,Fuzhou University,Fuzhou 350002,China
  • Received:2007-12-25 Revised:2008-03-03 Online:2009-01-01 Published:2009-01-01
  • Contact: CHENG Guo-qing

摘要: 提出网格相对密度的概念和边界点提取技术,在此基础上给出了一种多密度聚类算法。该算法使用网格相对密度识别具有不同密度聚簇的相对高密度网格单元,聚类时从相对高密度网格单元开始逐步扩展生成聚簇。实验结果表明,算法能有效地识别不同形状、不同密度的聚簇并对噪声数据不敏感,具有聚类精度高等优点。

关键词: 聚类, 多密度, 网格, 网格相对密度

Abstract: The concept of grid relative density(GRD) and boundary point extraction technique are proposed.Based on that,a clustering algorithm based on GRD for multi-density data sets(CM_GRD) is developed.In CM_GRD,GRD is used to identify relatively high density grid cells in clusters having different denstities.A clustering starts with one of relatively high density grid cells and then extends to its neighbors to form a cluster.Experiment results show that the proposed algorithm can discover clusters with various shapes and different densities and is insensitive to noise data.

Key words: clustering, multi-density, grid, grid relative density