多类分类问题是我们经常遇到的问题,常用的方法是将多类问题转化为若干个二类问题,然后利用二类支持向量机(support vector machine,SVM)进行分类,如一对余SVM,一对一SVM,决策树SVM等.在这些方法中,大都没有考虑所生成的多个分类器之间的可靠性和重要性问题.为了改进这一点,本文以一对余SVM为例,提出了两种基于可靠性测度的多类分类算法,算法的思想可用于一对一SVM,决策树SVM等其他多种分类器中.为了检验所提算法的有效性,本文进行了比较试验,实验结果表明所提算法不仅提高了分类准确度,而且具有更为广泛的推广能力. Many kinds of classification problems are the problems we often encounter. The commonly used method is to transform many kinds of problems into several kinds of second-class problems, and then classify them by using two kinds of support vector machines (SVMs) One-on-one SVM, decision tree SVM, etc. Most of these methods do not consider the reliability and importance of the generated multiple classifiers.In order to improve this, this paper takes a pair of SVM as an example, Two kinds of classification algorithms based on reliability measure are proposed, and the idea of ​​the algorithm can be used in many other classifiers such as one-to-one SVM and decision tree SVM. In order to test the validity of the proposed algorithm, The results show that the proposed algorithm not only improves the classification accuracy, but also has a broader promotion ability.