Hybrid and Ensemble Learning Framework for Accurate Classification of Imbalanced Data

This study evaluates advanced machine learning models for classification in imbalanced datasets. Two hybrid models, SVM–MLP and Autoencoder–XGBoost, are proposed and compared with LightGBM and XGBoost using MCC, Balanced Accuracy, and ROC–AUC. The hybrid models achieve superior performance, with SVM–MLP providing the most balanced and reliable predictions, while Autoencoder–XGBoost effectively handles class imbalance. The results demonstrate that hybrid learning frameworks combining feature extraction and powerful classifiers significantly improve classification performance in imbalanced datasets.