论文部分内容阅读
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。
Based on binary tree support vector machine multi-class classification algorithm in the current multi-class classification algorithm overall performance is better, but there are still classification accuracy and classification efficiency is not high. To solve these problems, this paper proposes a new multi-class classification algorithm based on Euclidean distance distance tree (DBT-SVM), which considers the Euclidean distance between two types of recent samples, The influence of the Euclidean distance of the center on the classification and the separation of the most easily separated classes are given priority. The experimental results on the UCI standard dataset show that the proposed algorithm is effective.