一种SVM多分类算法

A multi-class SVM classification algorithm

为了使用支持向量机(SVM)算法进行多类分类,在SVM二分类基础上,提出使用排序算法中冒泡排序的思想进行SVM多类别数据分类。使用该方法在选取的UCI数据集进行实验,结果表明,在保证较高正确率的情况下,相对传统一对一的多分类方法,该方法较大幅地减少了分类时间,是一种应用性较强的SVM多类分类方法。

To classify multi-class data with support vector machine（ SVM） algorithm,this paper proposes a new multi-class classification algorithm based on base SVM. This algorithm uses the thought of bubbling sort in sort algorithms. The experiment results on UCI datasets with this algorithm demonstrated that on the basis of high precision,it decreases the classification time significantlyrelative to one-against-one SVM multi-class classification algorithm. It is a highly practical multi-class SVM algorithm.