A Review of Hybrid Sampling and Ensemble Learning Techniques for Imbalanced Data Classification

Class imbalance is a significant challenge in machine learning, where the unequal distribution of classes often leads to biased prediction models and poor minority class recognition. This issue is commonly observed in critical domains such as healthcare, cybersecurity, fraud detection, and financial analysis, where accurate identification of rare events is essential Over the years, several techniques have been developed to address this problem, including data-level methods, algorithm-level approaches, ensemble learning, and hybrid strategies However, many existing methods still suffer from limitations related to scalability, overfitting, computational complexity, and limited adaptability to dynamic datasets. This paper presents a comprehensive review of recent hybrid learning techniques used for imbalanced data classification. The study analyzes commonly adopted sampling approaches such as SMOTE, ADASYN, Tomek Links, and Borderline-SMOTE along with ensemble techniques including Random Forest, AdaBoost, Gradient Boosting, and deep learning-based hybrid models. Furthermore, the paper discusses the strengths and limitations of existing approaches by comparing their performance using evaluation metrics such as Recall, F1-score, G-mean, and AUC Based on the analysis, several research gaps are identified, particularly in the areas of adaptive learning, scalability, and explainability Finally, this review highlights the need for intelligent and adaptive hybrid frameworks capable of handling evolving real-world imbalanced datasets efficiently. The findings of this study are expected to support researchers in developing robust and reliable machine learning models for future imbalance learning applications.