基于二叉树的半监督三分类光滑支持向量机研究

Study on Semi-supervised Triclassification Smooth Support Vector Machine Based on Binary Tree

文章对半监督三分类光滑支持向量机进行研究。首先,针对二叉树结构的分类顺序对分类精度影响的问题,提出采用基于类间相似方向数生成偏二叉树支持向量机的方法判别分类顺序。其次,将此判别方法应用于半监督三分类中,提出了基于类间相似方向数的半监督三分类光滑支持向量机算法。最后,进行五组数值试验。结果表明,与全监督三分类算法相比,本文算法得到的准确率相对要高,且能在已标记样本比例较少的情况下得到较高的准确率,从而表明了本文的半监督二叉树三分类光滑支持向量机具有优越性。

This paper makes a study on the semi-supervised tri-classification smooth support vector machine. Firstly,aiming at the influence of classification order of binary tree structure on classification accuracy,this paper proposes a method of generating partial binary tree support vector machine based on the number of similar directions between classes. Secondly,the paper applies this method to semi-supervised tri-classification to present a semi-supervised tri-classification smooth support vector machine algorithm based on the number of similar directions between classes. Finally,five sets of numerical experiments are carried out. The results show that compared with the fully supervised tri-classification algorithm,the proposed algorithm in the paper has relatively higher accuracy,and can get high accuracy when the proportion of marked samples is small,thus indicating that the proposed semi-supervised binary tree tri-classification smooth support vector machine has advantages.