针对不平衡数据集的维数约简方法

Implementing Dimension Reduction on Unbalanced Data sets

针对不平衡数据的分类问题,本文提出了一种新的方法,将特征选择应用在不平衡数据集中,首先对数据集进行预处理,然后从特征选择的角度出发,选择具有较强能力代表数据集的特征,简化数据的同时也提高了分类性能。通过实验表明,该方法能够有效地提高分类精度。

This paper proposes a new method for imbalanced data classification. After the unbalanced data set is preprocessed by implementing feature selection, some features with strong data representative capabilities are left, and classifiers are constructed on the preprocessed dataset. Experimental results show this approach improves the classification accuracy for unbalanced datasets.