Combining SMOTE and OVA with Deep Learning and Ensemble Classifiers for Multiclass Imbalanced

The classification of real-world problems alwaysconsists imbalanced and multiclass datasets. A dataset having unbalanced andmultiple classes will have an impact on the pattern modeland accuracy, which be decreased. Hence,oversampling method keeps class balanced avoids theoverfitting problem. purposes study were to handle multiclassimbalanced datasets improve effectivenessof model. This proposed a hybrid bycombining Synthetic Minority Oversampling Technique (SMOTE) One-Versus-All(OVA) with deep learning ensemble classifiers; stacking random forestalgorithms for data handling. Datasets consisting ofdifferent numbers imbalances are gained from UCI MachineLearning Repository. research outputs illustrated that presented methodacquired best accuracy value at 98.51% when classifierwas used evaluate model performance in new-thyroiddataset. using algorithm received higheraccuracy rate than other methods car, pageblocks, Ecolidatasets. In addition, highest ofclassification 98.47% dermatology where forest isused as classifier.