Research of Imbalanced Data Classification in Data Mining

Classification is one of the most important research contents in data mining and traditional classification methods are relatively mature, when dealing with well-balanced data they can make good performances. But in real world the data is usually imbalanced, that is, most of the data are in majority class and little data are in minority class. Imbalanced data set cause the deduction of the precision of the minority class samples, when it is classified by traditional algorithm, which can tend to favor the more class samples. Making researches on imbalanced datasets are quite important. In order to help readers to have a clear idea of the currently proposed and future work data classification, in view of imbalanced data progress, this paper introduced three developed methods: data level, algorithmic level and developed methods that were the performance evaluation of imbalanced data classification. We are very glad to receive the valuable reference provided by the academics that interested in this field.