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

Abstract

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