试谈分类用于多分类不足最小二乘支持向量分类—回归机

论文导读：分类问题；三分类问题；最小二乘支持向量机；分类回归机；一对一对多方法:A英文标题Leastsquaresupportvectorclassificationregressionmachineformulticlassificationproblems英文作者名ZHAIJia,HUYiqing*,XUEr英文地址(SchoolofMathemati
文章编号:10019081(2013)07189404
doi:10.11772/j.issn.1001908

1.201

3.07.1894

摘 要： 基于支持向量机(SVM)的三分类方法是处理多分类问题的一类方法。提出了最小二乘支持向量分类回归机(LSSVCR)算法，通过最小二乘目标函数充分考虑所有样本点对分类的影响，使得训练集中即使有个别样本点被标错类别，对分类结果也不会产生太大的影响，从而提高分类的准确性。该方法能够提高分类的准确率和分类速度，同时算法对于不同类别间样本数目差异较大的情况也有很好的分类效果。数值实验结果表明所提算法是可行的，且与已有的三分类算法相比在分类准确性上平均提高了

2.57%，在运算速度上也有了较大的提高。

关键词：多分类问题；三分类问题；最小二乘支持向量机；分类回归机；一对一对多方法
:A
英文标题
Least square support vector classificationregression machine for multiclassification problems


英文作者名
ZHAI Jia, HU Yiqing*, XU Er
英文地址(
School of Mathematics and Physics, University of Science and Technology Beijing, Beijing 100083, China英文摘要)
Abstract：
Triclass classification method based on Support Vector Machine (SVM) is a kind of method for solving multiclass classification problems. Least Square Support Vector ClassificationRegression (LSSVCR) was proposed, which considered the effects of all the sample points by using least squares objective function. Even if there were wrongly marked sample points in the training set, the result would not be affected largely by them. LSSVCR was more accurate and faster, and it was efficient for the problems that there are large differences among the number of sample points in different classes. The numerical experiments show that the proposed method raises the accuracy by 2.57% on erage compared to the existing triclassification methods.

Triclass classification method based on Support Vector Machine (SVM) is a kind of method so源于：论文例文www.7ctime.com
lving multiclass classification problems. Least Square Support Vector ClassificationRegression (LSSVCR) was proposed, which considered the effects of all the sample points by using least squares objective function. Even if there were wr论文导读：
ongly marked sample points in the training set, the result would not be affected largely by them. LSSVCR was more accurate and faster, and it was efficient for the problems which had large differences in the number of sample points in different classes. The numerical experiments show that the proposed method raises the accuracy by 2.57% comparing to the existed triclassification methods.
英文关键词Key words:
multiclass classification problem; triclass classification problem; Least Square Support Vector Machine (LSSVM); classificationregression machine; one versus one versus rest (1v1vr) method



0 引言
分类问题广泛存在于生产生活及许多研究领域中。二分类问题是研究较多也较为成熟的分类问题，支持向量机(Support Vector Machine, SVM)［1］46-59则是解决二分类问题常用的方法之一。
实际分类问题中所面对的更多的是多分类问题。目前已经有众多学者研究解决多分类问题。最常用的方法是将多分类转化为一系列二分类问题［2］273-297，最终通过二分类结果的组合得到多分类的结果。此类方法包括一对多(One versus rest，1vr)［3］、一对一(One versus one，1v1)［4］、纠错输出编码方法［5］等。以k分类问题为例，一对多算法将第i类问题(i=1,2,…,k)看作是一类，而将剩余的k－1类问题看作另一类，从而将多分类转化为k个二分类问题。一对一算法则对全部或部分类型对{(i, j)|1≤i