论文部分内容阅读
k-近邻分类是一种流行且成功的非参数分类方法,但其分类性能由于离群点的存在而受到损害.为克服离群点对分类性能的不利影响,提出了一个k-近邻分类的变形和一个基于局部均值向量与类均值向量的近邻分类方法.该方法利用了未分类样本在每个训练类中k个近邻的局部均值的信息和整体均值的知识,不仅能够克服离群点对分类性能的影响,而且取得了比传统的k-近邻分类一致好的分类性能.
k-nearest neighbor classification is a popular and successful non-parametric classification method, but its classification performance is impaired by the existence of outliers.In order to overcome the adverse effect of outliers on the classification performance, a k-nearest neighbor classification Deformation and a neighborhood classification method based on local mean vector and class mean vector.Using the local mean information and global mean of k nearest neighbors in each training class, this method can not only overcome the outlier pairs Classification performance, but also achieved better classification performance than the traditional k-nearest neighbor classification.