An in-depth performance analysis of the oversampling techniques for high-class imbalanced dataset

Class imbalance occurs when the distribution of classes between majority and minority is not same. The data on imbalanced may vary from mild to severe. effect high-class affect overall classification accuracy since model most likely predict that fall within class. Such a will give biased results, performance predictions for class often have no impact model. use oversampling technique one way deal with imbalance, but only few are used solve imbalance. This study aims an in-depth analysis techniques address problem. addition balance each class’s provide unbiased evaluation results in modeling. We compared Random Oversampling (ROS), ADASYN, SMOTE, Borderline-SMOTE techniques. All be combined machine learning methods such as Forest, Logistic Regression, k-Nearest Neighbor (KNN). test show Forest gives best value 0.9997, 0.9474 precision, 0.8571 recall, 0.9000 F1-score, 0.9388 ROC-AUC, 0.8581 PRAUC technique.