数据不平衡分类研究综述

Survey of Classification with Imbalanced Data

在分类领域中,当数据集不平衡时,传统的分类算法和评估指标都不能很好地对数据分类。因此,多年来不少学者针对这一领域进行研究。主要分为三大类,即抽样方法、代价敏感方法、集成方法。同时针对这个领域枚举一些评估指标。

In classification field, when the data is imbalanced, the traditional classification algorithms and evaluation criteria are not good for it. So,a lot of researchers study it recent years. Mainly divides into three categories, such as resample technique, cost-sensitive learning and ensemble techniques. At the same time, puts forward some new standards to evaluate the algorithms in this field.