Hybrid and Ensemble Machine Learning Models for Effective Imbalanced Data Classification

This study explores the effectiveness of advanced machine learning models in addressing classification challenges within imbalanced datasets. In this paper, two models are developed, specifically Support Vector Machine (SVM) with Multi-Layer Perceptron (MLP) and Autoencoder + Extreme Gradient Boosting (XGBoost), while two models, Light Gradient Boosting Machine (LightGBM) and XGBoost, are implemented for comparison.These models are evaluated using robust metrics such as Matthews Correlation Coefficient (MCC), Balanced Accuracy, and Receiver Operating Characteristic Area Under the Curve (ROC AUC) Score to ensure a comprehensive assessment. The results reveal the exceptional performance of the developed hybrid models, with the SVM with MLP model emerging as the most reliable due to its strong balance between precision and recall. The Autoencoder + XGBoost model closely follows, excelling in handling imbalanced data and delivering consistent results. Among the implemented models, LightGBM showcases robust classification capabilities, offering a scalable and effective alternative, while XGBoost, although reliable, demonstrates limitations in handling complex class distributions compared to the developed approaches. This analysis underscores the importance of leveraging hybrid architectures for achieving high-performance classification in data-intensive and imbalanced scenarios. By combining feature extraction and advanced classification techniques, these models not only improve predictive accuracy but also ensure better handling of class imbalance.