摘要
该文提出一种具有N-S磁极效应的最大间隔模糊分类器(MPMMFC)。该方法寻求一个具有N-S磁极效应的最优超平面,使得一类样本受磁极吸引离超平面尽可能近,另一类样本受磁极排斥离超平面尽可能远。针对传统支持向量机面临的对噪声和野点敏感问题,引入模糊技术来降低噪声和野点对分类的影响,从而进一步提高泛化性能和分类效率。通过人工数据集和实际数据集上的实验,证明了MPMMFC的有效性。
Inspired by space geometry and magnetic pole effect theory, a maximum margin fuzzy classifier with N-S magnetic pole（MPMMFC） is proposed in this paper. The main idea is to find an optimal hyperplane based on N-S magnetic pole effect in order to ensure that the distance between one class and the hyperplane is much closer due to pole attractive and the distance between the other class and the hyperplane is much greater due to repulsion. Moreover, due to the traditional support vector machine（SVM） sensitive to noises and outliers, a fuzzy technology is introduced in this paper to reduce the influence of noises and outliers, and the classification efficiencies and generalization performance are improved further. Experimental results on the synthetic datasets and UCI datasets show that the proposed approaches are effective.
出处
《电子科技大学学报》
EI
CAS
CSCD
北大核心
2016年第2期227-232,239,共7页
Journal of University of Electronic Science and Technology of China
基金
国家社科基金后期资助项目(15FTQ008)
关键词
模糊技术
核方法
磁极效应
模式分类
fuzzy technology
kernel method
magnetic pole
pattern classification
