一种新型基于二叉树的支持向量机多类分类方法

A new SVM multiclass classification method based on binary tree

采用聚类分析中的类距离思想,通过类的重心,计算该类与其他类间的平均距离,提出了一种新的二叉树生成算法.在算法中,利用对称矩阵的特点,简化计算,并对先分离出来的类距离进行有效舍弃.试验结果表明该算法具有一定的有效性.

Using class distance of clustering and computing the average distance of class through the barycenter of class, an improved multiclass SVM based on binary tree was proposed. In the algorithm, the computation can be predigested using the symmetric matrix, and the distance of separated calss can be effectively abnegated. The experiment results show that the muhiclass SVM method is suitable for practical use.