计算机应用

• 人工智能 • 上一篇    下一篇

基于Fisher准则和特征聚类的特征选择

王飒 郑链   

  1. 北京理工大学宇航科学技术学院 北京理工大学宇航科学技术学院
  • 收稿日期:2007-05-11 修回日期:1900-01-01 发布日期:2007-11-01 出版日期:2007-11-01
  • 通讯作者: 王飒

Feature selection method based on Fisher criterion and feature clustering

Sa Wang Lian Zheng   

  • Received:2007-05-11 Revised:1900-01-01 Online:2007-11-01 Published:2007-11-01
  • Contact: Sa Wang

摘要: 特征选择是机器学习和模式识别等领域的重要问题之一。针对高维数据,提出了一种基于Fisher准则和特征聚类的特征选择方法。首先基于Fisher准则,预选出鉴别性能较强的特征子集,然后在预选所得到的特征子集上对特征进行分层聚类,从而最终达到去除不相关和冗余特征的目的。实验结果表明该方法是一种有效的特征选择方法。

关键词: 特征选择, Fisher准则, 特征聚类

Abstract: Feature selection is one of the important issues in the machine learning and pattern recognition. For highdimensional data, the new feature selection method based on Fisher criterion and feature clustering was proposed. Firstly, the features that have more discrimination information based on Fisher criterion were selected. Then hierarchical clustering in the preselected subset was adopted. Finally the irrelevant and redundant features were removed. The experimental results show that the proposed algorithm is an effective method for feature selection.

Key words: feature selection, Fisher criterion, feature clustering