非平衡数据集分类方法探讨

Discussion of Classification for Imbalanced Data Sets

由于数据集中类分布极不平衡,很多分类算法在非平衡数据集上失效,而非平衡数据集中占少数的类在现实生活中通常具有显著意义,因此如何提高非平衡数据集中少数类的分类性能成为近年来研究的热点。详细讨论了非平衡数据集分类问题的本质、影响非平衡数据集分类的因素、非平衡数据集分类通常采用的方法、常用的评估标准以及该问题中存在的问题与挑战。

Because of imbalanced class distribution,most classifiers lose efficiency with it.In fact the rarely occurring class in imbalanced datasets shows statistical significance.The problem of learning from imbalanced datasets has attracted growing attention in recent years.The paper provided a comprehensive review of the classification of imbalanced datasets,the nature of the problem,the factor which affected the problem,the current assessment metrics used to evalua-te learning performance,as well as the opportunities and challenges in the learning from imbalanced data.