计算机工程 ›› 2011, Vol. 37 ›› Issue (5): 224-226.doi: 10.3969/j.issn.1000-3428.2011.05.076

• 人工智能及识别技术 • 上一篇    下一篇

子空间可能性聚类机制研究

关 庆,邓赵红,王士同   

  1. (江南大学信息工程学院,江苏 无锡 214122)
  • 出版日期:2011-03-05 发布日期:2012-10-31
  • 作者简介:关 庆(1986-),女,硕士研究生,主研方向:聚类算法,智能计算;邓赵红,副教授、博士;王士同,教授、博士生导师
  • 基金项目:
    国家自然科学基金资助项目(60903100);江苏省自然科学基金资助项目(BK2009067)

Research on Subspace Possibilistic Clustering Mechanism

GUAN Qing, DENG Zhao-hong, WANG Shi-tong   

  1. (School of Information Engineering, Jiangnan University, Wuxi 214122, China)
  • Online:2011-03-05 Published:2012-10-31

摘要: 可能性C-均值(PCM)聚类作为经典的基于原型的聚类方法,在处理高维数据集时性能骤降,无法检测出高维空间中嵌入的有效子空间。针对此不足,在PCM基础上引入子空间聚类机制,提出子空间可能性聚类算法SPC。该方法保留了PCM方法的优点,且对高维数据具有较好的适应性,能够有效检测各类所处的子空间。仿真实验验证了SPC算法的有效性。

关键词: 高维数据, 子空间聚类, 特征加权, 可能性聚类

Abstract: The obvious shortcomings of Possibilistic C-Means(PCM) algorithm is that the performance will be significantly reduced for high dimensional data sets and it can not effectively identify the useful subspace embedded in the high dimensional space. In order to overcome the weakness, the subspace clustering mechanism is introduced and the Subspace Possibilistic Clustering(SPC) algorithm is presented. It not only has the advantages of PCM algorithm but also has the characteristic of the classic subspace clustering algorithms. Namely, it has good adaptability to high dimensional data, and can detect the subspaces for each cluster effectively. Simulation experiments with synthetic and real data sets demonstrate the effectiveness and the merits of SPC.

Key words: high dimensional data, subspace clustering, feature weighting, possibilistic clustering

中图分类号: