Abstract: To overcome the disadvantages of Unclassifiable Region (UR) in conventional Multi-classification Support Vector Machine (MSVM), and increase the classification capacity and generalization ability of MSVM, Euclidean Distance Method (EDM) was presented. EDM computed the distances between the sample in UR and every class center directly, and then selected the class with the least Euclidean distance for the sample. The experimental results on benchmark datasets show that EDM eliminates the UR in conventional MSVM and improves the classification capacity and generalization ability of MSVM effectively.

Key words: Euclidean distance, Unclassifiable Region (UR), multi-classification, Support Vector Machine (SVM)

摘要： 为克服传统多分类支持向量机中存在的拒识区域问题，提高算法的分类性能和泛化能力，提出一种基于欧氏距离的拒识区域解决方案。该方法直接计算落入拒识区域中的样本点到每类中心的欧氏距离，然后选择较小的欧氏距离对应的类为样本的所属类。基于标准数据集的实验结果表明，欧氏距离法实现了零拒识，有效提高了算法的分类性能和泛化能力。

关键词: 欧氏距离, 拒识区域, 多分类, 支持向量机