一种基于半监督学习的非平衡分类算法

An Imbalanced Classification Algorithm Based on Semi-supervised Learning

半监督学习是近年来流行的一种机器学习方法,主要是解决在分类问题中,未标注样例充足但有标注样例短缺的问题.但当前的大多数半监督学习算法都假定样例数据是平衡的,这在现实世界中很多情况下是不真实的.提出一种新的基于半监督学习的非平衡分类算法,通过随机动态生成样例特征子空间,有效地处理了样例数据的不平衡问题.在4个相关数据集上的实验验证了本方法的有效性.

In many real-world applications plenty of unlabeled instances are available but the number of labeled instances is limited. In order to solve this problem, various semi-supervised learning methods have been proposed recently. However, most existing studies assume the balanee between negative and