基于样本投影分布的平衡不平衡数据集分类

Balanced and imbalanced data sets classification based on sample projection distribution

提出一种平衡不平衡数据集统一分类方法,首先得到训练样本基于支持向量机(SVM)超平面法线方向上的投影;再借助支持向量数据描述(SVDD)对训练样本投影分布进行描述;测试样本在此基础上实现分类。平衡或不平衡数据集都可采用相同的方法进行分类。实验表明该方法能够同时对平衡或不平衡数据集进行有效的分类。

This paper proposed a new unified classification method for balanced and imbalanced data sets. Firstly, obtained training sample projection based on SVM hyperplane normal direction, secondly, acquired the description for training sample projection distribution by means of SVDD, at last, realized classification for test sample based on the distribution description. The proposed method unified classification form for balanced and imbalanced data sets. The experiments show that the method has good classification appearance for balanced and imbalanced data sets.