Over-sampling via under-sampling in strongly imbalanced data

Classification of imbalanced datasets is an important challenge in machine learning. This investigation analysed the effect of ratio imbalance and the selected classifier on the application of several re-sampling strategies to deal with imbalanced datasets. We applied two different classifiers J48 and Naive Bayes, four re-sampling algorithms Org, SMOTE, Borderline SMOTE, OSS and NCL approaches and four performance assessment measures TPrate, TNrate, Gmean and AUC on 13 sets of real data. Our experimental results show that, whenever, datasets are strongly imbalanced, over-sampling methods are more efficient in compare with under-sampling methods. Moreover, our results indicate that, when dealing with imbalanced data with any level, applying re-sampling techniques is preferred. Further, the results indicate that the classifier has very poor influence on the effectiveness of the re-sampling strategies.