计算机工程与应用 ›› 2012, Vol. 48 ›› Issue (6): 146-150.

• 图形、图像、模式识别 • 上一篇    下一篇

一种用于分类的改进Boosting算法

刘 凯1,王正群2   

  1. 1.江苏畜牧兽医职业技术学院,江苏 泰州 225300
    2.扬州大学 信息工程学院,江苏 扬州 225009
  • 收稿日期:1900-01-01 修回日期:1900-01-01 出版日期:2012-02-21 发布日期:2012-02-21

Improved Boosting algorithm for classification problems

LIU Kai1, WANG Zhengqun2   

  1. 1.Jiangsu Animal Husbandry and Veterinary College, Taizhou, Jiangsu 225300, China
    2.School of Information Engineering, Yangzhou University, Yangzhou, Jiangsu 225009, China
  • Received:1900-01-01 Revised:1900-01-01 Online:2012-02-21 Published:2012-02-21

摘要: 提出了一种新的Boosting算法LAdaBoost。LAdaBoost算法利用局部错误率更新样本被选用于训练下一个分类器的概率,当对一个新的样本进行分类时,考虑了该样本与其邻域内的每个训练样本的近似度;另外,提出了有效邻域的概念。根据不同的组合方法,得到了两种LAdaBoost算法,即LAdaBoost-1和LAdaBoost-2。在UCI上部分实验数据集的实验结果表明,LAdaBoost算法比AdaBoost和Bagging算法更有效,且鲁棒性更好。

关键词: 机器学习, Bagging算法, Boosting算法, 噪声

Abstract: A new Boosting algorithm named LAdaBoost is proposed, which utilizes a local error to update the probability that the instance is selected to be part of next classifier’s training set. When classifying a new instance, the similarity between the instance and each training instance in its neighborhood is taken into account. Furthermore, the concept of effective neighborhood is first given. According to different combination methods, it gets two LAdaBoost algorithms LAdaBoost-1 and LAdaBoost-2. The experimental results on several datasets available from the UCI repository demonstrate that LAdaBoost algorithms are more robust and efficient than AdaBoost and Bagging.

Key words: machine learning, Bagging, Boosting, noise