失衡数据集分类技术研究进展

Imbalanced data sets (IDS) is an objective existence data form in the field of data mining and machine learning, many areas of the real world has extensive application. This paper introduced the characteristics of imbalanced data sets, and detailed analysis of the imbalanced data classification problems faced form two aspects of data distribution and decision space. This paper also described the method to improve the classifier performance from the two aspects of data resampling and classification algorithm improving, introduced the main methods and strategy experts and scholars from various countries to solve the problem of IDS classification. Compared with the traditional classifier performance measure and introduced applicable measure to evaluate the IDS classification performance.