In recent decades, ensemble methods like bagging and boosting have seen widespread adoption in machine learning for their strong performance on classification tasks. These techniques have proven particularly effective at addressing the persistent challenge of class imbalance. However, while boosting methods have been the subject of extensive research, a systematic analysis and consolidation of bagging approaches for imbalanced data remains scarce. This paper addresses this gap by providing a comprehensive review and empirical analysis of bagging-based techniques tailored for imbalanced datasets. Furthermore, we note the absence of standardized, publicly available implementations for many of these algorithms. To bridge this gap, we introduce ImbBag, a Python package that provides a unified, scikit-learn-compatible library of key bagging methodologies for both binary and multi-class imbalanced problems. This package is designed as a practical tool for both researchers and practitioners. The source code and comprehensive documentation are available on GitHub at: https://github.com/yousefabdi/imbbag.
