A novel imbalanced data classification approach using both under and over sampling

The performance of the data classification has encountered a problem when distribution is imbalanced. This fact results in classifiers tend to majority class which most instances. One popular approaches balance dataset using over and under sampling methods. paper presents novel pre-processing technique that performs both algorithms for an imbalanced dataset. proposed method uses SMOTE algorithm increase minority class. Moreover, cluster-based approach performed decrease takes into consideration new size experimental on 10 datasets show suggested better comparison previous approaches.