<![CDATA[Data balancing methods in multi-class settings continue to evolve as the importance of balanced data conditions for classification analysis grows. However, limited studies have provided comprehensive empirical comparisons across both binary and multi-class imbalanced datasets. Data imbalance can affect model predictions, particularly by leading to inaccurate identification of minority classes. Therefore, this study aims to evaluate the effectiveness of the Synthetic Minority Over-sampling Technique (SMOTE) in improving classification performance. Three benchmark datasets from the UCI Machine Learning Repository—Breast Cancer, Ecoli, and Glass—were selected to represent imbalanced classification problems in both binary and multi-class settings. The proposed framework addresses class imbalance during data preprocessing using SMOTE. Each dataset is first divided into training and testing subsets. SMOTE is applied only to the training data to address class imbalance, while the test data is kept unchanged for evaluation. Then, the classification process is applied to the original (imbalanced) data and to the balanced data generated by SMOTE. The classifiers used in this study are SVM, a decision tree, and AdaBoost. The classification results are evaluated based on accuracy, sensitivity, and F1-score. The results show that the decision tree and AdaBoost improve classification performance under imbalanced data conditions. In particular, AdaBoost achieves the best overall performance in terms of prediction accuracy and class balance, demonstrating the effectiveness of combining SMOTE with ensemble methods for handling imbalanced datasets.]]>
Copyrights © 2026
