基于球结构支持向量机的多标签分类的主动学习

Active learning for multi-label classification based on sphere structured support vector machine

为了实现数据的多标签分类,减少多标签训练样本开销,将球结构支持向量机与主动学习方法结合用于多标签分类,依据球重叠区域样本距离差值度确定样本类别,分析多标签分类特性,采用样本近邻方法更新分类器。实验结果表明,该方法可以用较少的训练样本获得更有效的分类结果。

In order to implement the muhi-label classification of data and reduce the overload of muhi-label training samples, an algorithm combined with sphere structured Support Vector Machine （SVM） and active learning method was proposed. The labels of the samples in overlapping regions were determined according to distance difference value. The classification features of multi-label were analyzed. Then classifier was updated by closed neighbor method. The experimental results show that the method can achieve more efficient results using less training samples.