计算机工程与应用 ›› 2007, Vol. 43 ›› Issue (31): 23-24.

• 博士论坛 • 上一篇    下一篇

一种基于特征选择的不完整数据分类方法

陈景年1,2,黄厚宽1,田凤占1,薛小平3   

  1. 1.北京交通大学 计算机与信息技术学院,北京 100044
    2.山东财政学院 信息与计算科学系,济南 250014
    3.北京交通大学 电子信息工程学院,北京 100044
  • 收稿日期:1900-01-01 修回日期:1900-01-01 出版日期:2007-11-01 发布日期:2007-11-01
  • 通讯作者: 陈景年

Classification method for incomplete data based on feature selection

CHEN Jing-nian1,2,HUANG Hou-kuan1,TIAN Feng-zhan1,XUE Xiao-ping3   

  1. 1.School of Computer and Information Technology,Beijing Jiaotong University,Beijing 100044,China
    2.Department of Information and Computing Science,Shandong University of Finance,Ji’nan 250014,China
    3.School of Electronics and Information Engineering,Beijing Jiaotong University,Beijing 100044,China
  • Received:1900-01-01 Revised:1900-01-01 Online:2007-11-01 Published:2007-11-01
  • Contact: CHEN Jing-nian

摘要: 特征选择(也称作属性选择)是简化数据表达形式,降低存储要求,提高分类精度和效率的重要途径。实际中遇到的大量的数据集包含着不完整数据。对于不完整数据,构造选择性分类器同样也可以降低存储要求,提高分类精度和效率。因此,对用于不完整数据的选择性分类器的研究是一项重要的研究课题。有鉴于此,提出了一种用于不完整数据的选择性贝叶斯分类器。在12个标准的不完整数据集上的实验结果表明,给出的选择性分类器不仅分类准确率显著高于非常有效地用于不完整数据的RBC分类器,而且分类性能更加稳定。

Abstract: Feature selection is an important policy to simplify data,reduce necessary memory and improve the accuracy and efficiency of classification.Data are often incomplete because of various kinds of reasons.For incomplete data,methods of constructing selective classifiers can also reduce necessary memory and improve the accuracy and efficiency of classification.So developing selective classifiers for incomplete data is an important problem.In this paper a method of constructing selective Bayes classifiers from incomplete data is presented.Experiments on twelve benchmark incomplete data sets show that not only is the classification accuracy of the selective classifier proposed much higher than that of the very efficient RBC classifier,but also its performance is more robust.