电信科学

• • 上一篇    下一篇

基于密度与动态阈值的任意形状聚类挖掘算法研究

陈沛帅,琚春华   

  1. 浙江工商大学信息学院;浙江工商大学信息学院
  • 出版日期:2012-01-15 发布日期:2012-01-15
  • 基金资助:
    国家自然科学基金资助项目(No.71071141,No.71001088,No.61070059);浙江省科技计划基金资助项目(No.2009C33015)

The Cluster Algorithm Research Based on Dynamic Variable Threshold and Density

Chen Peishuai and Ju Chunhua   

  1. Informantion College,Zhejiang Gongshang Univeristy;Informantion College,Zhejiang Gongshang Univeristy
  • Online:2012-01-15 Published:2012-01-15

摘要: 本文分析了数据聚类算法BIRCH的不足之处,提出了一种基于密度与动态阈值的任意形状聚类挖掘算法——DVTD算法,它结合密度和阈值双重参数,并根据数据集内在特征,动态改变阈值T,既可以控制CF树的大小,也能利用不同球形聚类逼近任意形状的数据聚类。实验结果表明,它的算法复杂度与BIRCH相当,并大大降低了CF的大小,对任意形状的聚类效果可以达到与DBSCAN相近的效果。

Abstract: BIRCH and DBSCAN are popular data cluster algorithms.However,they insist some insufficiency.This paper introduces a new algorithm DVTD(the cluster algorithm based on dynamic variable threshold and density) to solve these problems.In the experiment,it is shown that DVTD is better than BIRCH and it can get almost same result to arbitrary shapes data cluster as DBSCAN.

No Suggested Reading articles found!