基于类心距离的模糊支持向量数据描述

Fuzzy Support Vector Data Description with Centers of Classes Distance

针对传统的支持向量数据描述模型忽略了样本分布的重要性,提出了基于类心距离的模糊支持向量数据描述算法,并将其应用在UCI机器学习数据库的二分类和多分类数据集中。该算法利用样本到两类中心距离的比值赋予样本权重,增大贡献度大的样本的权重,降低贡献度小的样本的权重,突出样本之间的差异性,从而提高了算法的分类效果。实验表明,该算法具有比传统支持向量数据描述更好的学习能力和分类效果。

Support vector data description（SVDD）ignores the importance of sample distribution.This paper proposed a new method called fuzzy SVDD with centers of classes distance,and it had been applied on the UCI data sets.The algorithm uses ratio of the distance of sample to the centers of two classes to give each sample a weight.The important samples＇ weights should be increased and the others should be not,which can highlight the difference of samples.The results show that our algorithm has better performance than SVDD.