一种新的模糊支持向量机多分类算法

New multiclassification algorithm based on fuzzy support vector machines

在模糊多分类问题中,由于训练样本在训练过程中所起的作用不同,对所有数据包括异常数据赋予一个隶属度。针对模糊支持向量机(fuzzy support vectormachines,FSVM)的第一种形式,引入类中心的概念,结合一对多1-a-a(one-against-all)组合分类方法,提出了一种基于一对多组合的模糊支持向量机多分类算法,并与1-a-1(one-against-one)组合和1-a-a组合的分类算法比较。数值实验表明,该算法是有效的,有较高的分类准确率,有更好的泛化能力。

In the fuzzy multiclassification problem, gave a degree of membership to all the data including abnormal data as the training samples played different affections in the training procession. Facing to the first form of fuzzy support vector machines, used the concept of the class center. Considered with the one-against-all association assorting method,put out a new fuzzy support vector machines multiclassification model based on one-against-all association, and compared with one-against-one and one-against-all association assorting method. The numerical test has improved that the algorithm is effective, and it has higher accurate rate of classification, also better ability of generalization.