摘要
采用聚类分析中的类距离思想,在特征空间中,计算各类别间的最短距离,以最短距离计算该类与其它类的平均距离,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,同时实现了对先分离出来的类的类距离的有效舍弃,实验结果表明该算法具有一定的优越性。
This paper proposes an improved multiclass SVM based on binary tree by using class distance of clustering.To compute the average distance with the minimum distance between classes in feature space.Use the characteristics of the symmetric matrix to simplify the computation and abnegate the distance of ahead separated classes.The experiment results showed that the multiclass SVM method was suitable for practical use.
出处
《微型电脑应用》
2009年第3期46-47,6,共2页
Microcomputer Applications
关键词
文本分类
支持向量机
二叉树
聚类
Text categorization
Support vector machines
Binary tree
Clustering
