基于类别信息的分类器集成方法Cagging  

Cagging: Classifiers ensemble method based on class information

作　　者：刘风[1] 王正群[1] 陈广花[1] 

机构地区：[1]扬州大学信息工程学院,江苏扬州225009

出　　处：《计算机应用》2008年第S2期74-76,82,共4页journal of Computer Applications

基　　金：国家自然科学基金资助项目(60774017);江苏省高校自然科学基金资助项目(07KJB520133)

摘　　要：提出一种基于类别信息的分类器集成方法Cagging。基于类别信息重复选择样本生成基本分类器的训练集,增强了基本分类器之间的差异性;利用基本分类器对不同模式类的分类能力为每个基本分类器设置一组权重,使用权重对各分类器输出结果进行加权决策,较好地利用了各个基本分类器之间的差异性。在人脸图像库ORL上的实验验证了Cagging的有效性。此外,Cagging方法的基本分类器生成方式适合于通过增量学习生成集成分类器,扩展Cagging设计了基于增量学习的分类器集成方法Cagging-I,实验验证了它的有效性。A classifier ensemble method Cagging based on class information was proposed.Training sets of each classifier generated through selecting samples repeatedly based on class information enhanced the diverse of each classifier.The classify results of all classifiers were combined by voting with weight vector to use the diverse of each classifier better.This weight vector was set for each classifier according to its classification performance to each class.The experimental results on face database ORL verify Cagging's validity.Furthermore,the method to generate the classifier in Cagging can generate new ensemble classifier through incremental learning.So,a classifier ensemble method Cagging-I based on increment learning was designed by extending Cagging,and the experimental results verify its validity.

关 键 词：分类 分类器 类别信息 分类器集成 增量学习 

分 类 号：TP182[自动化与计算机技术—控制理论与控制工程]