基于不完全信息的One-class支持向量机

One-class support vector machine based on incomplete information

标准的单值支持向量(O ne-class SVM)机不能对含有不完全信息的输入样本进行学习分类。为此该文提出用区间数来对不完全输入信息进行描述,将不完全的信息输入扩展为区间向量形式,引入区间运算来取代原来分类函数中的运算,从而根据区间运算结果来对信息不完全的模式输入进行分类。使用该方法,在分类过程中能够充分利用区间表示的先验知识,同时也能够减少该过程中输入模式中的属性(特征)度量代价,理论分析和实验结果均表明该方法能最大程度地保证分类结果的一致性,是有效和可行的。

Standard one-class support vector machines one-class SVM cannot classify patterns containing incomplete information input. This paper presents an algorithm which uses interval number to represent the incomplete information, transforms the incomplete information input to interval vector form, and extends traditional operation to interval operation. The incomplete information input patterns are then classified according to the interval operation output. The attribute value prior knowledge represented by inter...