摘要
传统的支持向量机对噪声或野点是敏感的,针对这种情况,引入了模糊支持向量机,但模糊隶属度的确定是个难点。利用基于线性规划下的一类分类算法来确定模糊隶属度,根据不同输入样本对分类的贡献不同,赋予相应的隶属度,将噪声或野点与有效样本区分开。实验结果表明,模糊支持向量机比传统的支持向量机有更好的分类效果,能够削弱噪声或野点的影响。
As the traditional Support Vector Machines(SVM) is sensitive to the noises or outliers,Fuzzy Support Vector Machines is introduced,but the determination of fuzzy membership is a difficulty.One-class classification algorithm based on linear programming is used to determine fuzzy membership in this paper,then the corresponding membership is given according to different input data affects on the classification results.So this method effectively distinguishes between the noises or outliers and the valid samples.Experimental results indicate that Fuzzy Support Vector Machines yields better classification result than the traditional SVM,thus the effects of the noise or outliers can be diminished.
出处
《计算机工程与应用》
CSCD
北大核心
2008年第11期41-42,46,共3页
Computer Engineering and Applications
关键词
线性规划
模糊支持向量机
隶属度
linear programming
Fuzzy Support Vector Machines(FSVM)
membership