北京邮电大学学报

  • EI核心期刊

北京邮电大学学报 ›› 2002, Vol. 25 ›› Issue (2): 8-13.

• 学术论文 • 上一篇    下一篇

一种新的聚类算法:等密度线算法

赵艳厂,谢帆,宋俊德   

  1. 北京邮电大学电子工程学院, 北京 100876
  • 收稿日期:2001-06-25 出版日期:2002-03-10
  • 作者简介: 赵艳厂(1977-),男,博士研究生
  • 基金资助:
     

DILC: A Clustering Algorithm Based on Density-isoline

ZHAO Yan-chang, XIE Fan, SONG Jun-de   

  1. Electronic Engineering School, Beijing University of Posts and Telecommunications, Beijing 100876, China)
  • Received:2001-06-25 Online:2002-03-10
  • Supported by:
     

摘要: 提出了一种新的聚类算法:等密度线聚类算法。该算法从样本分布等密度线图的思想出发,从图中找出样本分布比较集中的区域,从而发现隐含在样本集中的类。等密度线聚类算法不需要输入任何参数,是一种无监督式聚类。它能够自动发现任意形状的类,并且能有效地排除噪声干扰。实验结果表明,等密度线聚类算法具有较快的聚类速度和较好的聚类效果。

关键词: 数据挖掘, 聚类, 等密度线聚类

Abstract: A new clustering algorithm, density-isoline clustering(DILC) algorithm is put forward in this paper. DILC starts from the density-isoline figure of samples, and finds relatively dense regions, which are clusters. DILC is capable of eliminating outliers and discovering clusters of various shapes. It is an unsupervisedclustering algorithm because it requires no interaction. The high accuracy andefficiency of DILC clustering algorithm are shown in our experiments.

Key words: data mining, clustering, density-isoline clustering

中图分类号: