基于二叉树支持向量机多类分类算法的研究  

Research on Multiclass Classification Algorithm Based on Binary Tree SVM

作　　者：吴恩英 吕佳[1] 

机构地区：[1]重庆师范大学计算机与信息科学学院,重庆401331

出　　处：《重庆师范大学学报（自然科学版）》2016年第3期102-106,共5页Journal of Chongqing Normal University:Natural Science

基　　金：重庆市自然科学基金(No.cstc2014jcyjA40011);重庆市教育委员会科学技术项目(No.KJ1400513)

摘　　要：基于二叉树的支持向量机多类分类算法虽然在目前现有的多类分类算法中总体性能较优,但是仍然存在分类精度和分类效率不高的问题。针对这些问题,提出了一种新的基于欧氏距离的二叉树支持向量机(Distance binary tree SVM,简称DBT-SVM)多类分类算法,该算法综合地考虑了两类最近样本的欧式距离、类中心的欧氏距离对分类的影响,并且使最容易分离的类能优先分离出来。通过在UCI标准数据集上进行实验验证,结果表明该算法行之有效。Multiclass classification algorithm based on binary tree SVM has good performance in many multiclass classification algorithms,algorithms;however,its classification accuracy and efficiency are still not high.To solve these problems,a new binary tree SVM multiclass classification algorithm based on euclideanEuclidean distance is proposed in this paper.The algorithm considers the effect of the euclideanEuclidean distances between the two nearest samples and two class centers.Moreover,the algorithm can give priority to those easiest separated classes.The experimental results on UCI benchmark datasets show that the presented algorithm is effective.

关 键 词：支持向量机 多类分类 二叉树 欧氏距离 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]