基于支持向量机的多类分类算法综述

An Overview of Multi-Class Algorithm Based on Support Vector Machine

摘要:

作为一种新兴的机器学习方法,基于统计学习理论的支持向量机,最初是用来解决二类分类问题的。对于实际中主要遇到的多类分类问题,目前常用的两大类改进推广方法为“分解—重组”法和“直接求解”法。文章对二类方法进行了介绍和分析,指出其优缺点和未来的改进方向。

Abstract:

As a new machine learning method, the support vector machine which is based on statistical learning theory, isused to solve binary classification problem originally. However, most of the classification problems in practice containmore than two classes, and there were two major types of methods to extend the binary SVM to multi-class SVM which are ‘Decomposition-Reorganization’method and‘Direct solving’method. In this paper, the two methods were introducedand analyzed and the advantages, disadvantages and the improvement direction in the future are pointed out.