基于主次原型超平面最接近支持向量机

基于广义特征值的最接近支持向量机GEPSVM是一种新的具有与SVM性能相当的两类分类方法,通过求解广义特征值来获得两个彼此不平行的拟合两类样本的超平面,其决策规则是将测试样本归为距其最近的超平面所在的类。然而,该规则在某些情形会导致较差的分类结果。对此,本文提出了在利用GEPSVM产生一个主原型超平面的基础上,再利用主原型超平面及它类样本的信息构造一个次原型超平面,形成一个由主次原型超平面共同决策的最接近支持向量机。该方法不仅简单且易于实现,而且具有较GEPSVM更优的分类性能。在UCI数据集上的实验验证了它的有效性。

The Proximal Support Vector Machine Based on the Primary and Secondary Prototypal Hyperplanes

A binary classifier termed as proximal support vector machine via generalized eigenvalues(GEPSVM) is proposed recently.It aims to obtain two nonparallel planes generated from their corresponding generalized eigenvalue problem and has an equivalent classification efficiency to SVM.In nature,GEPSVM attempts to fit two-class points using two planes.For an unknown sample,according to the decision rule of GEPSVM,it will be assigned to the closest planes.In fact,this rule,in many cases,may result in wrong classif...