📄 中文摘要
不平衡、重叠和噪声会降低数据质量，减少模型可靠性，并限制泛化能力。尽管在二分类中已广泛研究，这些问题在多类设置中仍然未得到充分探讨。复杂的类间关系使得少数类与多数类的结构不明确，传统聚类方法无法捕捉分布形状。仅依赖几何距离的方法可能会移除重要样本并生成低质量的合成数据，而二值化方法则局限于局部处理不平衡，忽视全局类间依赖性。在算法层面，集成方法难以整合弱分类器，导致鲁棒性有限。IMOVNO+ 提出了一个新的框架，旨在解决这些挑战，提升多类学习的效果。
IMOVNO+: 一种区域划分和元启发式集成框架用于不平衡多类学习
📄 English Summary
IMOVNO+: A Regional Partitioning and Meta-Heuristic Ensemble Framework for Imbalanced Multi-Class Learning
Class imbalance, overlap, and noise significantly degrade data quality, reduce model reliability, and limit generalization capabilities. While these issues have been extensively studied in binary classification, they remain underexplored in multi-class settings where complex inter-class relationships obscure minority-majority structures and traditional clustering fails to capture the distribution shape. Approaches relying solely on geometric distances risk discarding informative samples and generating low-quality synthetic data, while binarization methods address imbalance locally and overlook global inter-class dependencies. At the algorithmic level, ensemble methods struggle to integrate weak classifiers, resulting in limited robustness. IMOVNO+ proposes a novel framework aimed at addressing these challenges and enhancing the performance of multi-class learning.