Class Overlap Matters: Revisiting SOUP for Multi-Class Imbalanced Medical Data.

Saved in:

Bibliographic Details
Title:	Class Overlap Matters: Revisiting SOUP for Multi-Class Imbalanced Medical Data.
Authors:	Wang, Yujiang¹ wyjprogram@gmail.com, Mohd Rosli, Marshima¹ marshima@uitm.edu.my, Musa, Norzilah¹ norzilah@uitm.edu.my
Source:	IAENG International Journal of Computer Science. May2026, Vol. 53 Issue 5, p1615-1624. 10p.
Subjects:	Feature selection, Sampling methods, Machine learning, Data distribution, Data augmentation, Nosology
Abstract:	Multi-class imbalanced medical data classification is one of the most critical challenges in contemporary machine learning, as it reflects the diverse nature of disease conditions. While considerable efforts have been made to address class imbalance, recent studies have increasingly recognized that class overlap poses an even greater challenge to classification performance than the imbalance itself. A promising future direction to address the class overlap issue is to explore the complex inter-class relationships present in real-world datasets. In this context, the Similarity Oversampling and Undersampling Preprocessing (SOUP) method, which combines oversampling and undersampling techniques, has been considered a potential baseline for handling multi-class imbalance and provide useful insights for addressing class overlap. In this study, we revisit the SOUP method and examine its strengths and limitations in handling class overlap. Based on this analysis, we experimentally examine two potential modifications to the SOUP method: integrating feature selection and removing the undersampling component. Experiments on ten multi-class imbalanced medical datasets show that both lead to improved classification performance under class overlap conditions, providing a meaningful direction for future research. [ABSTRACT FROM AUTHOR]
	Copyright of IAENG International Journal of Computer Science is the property of International Association of Engineers (IAENG) and its content may not be copied or emailed to multiple sites without the copyright holder's express written permission. Additionally, content may not be used with any artificial intelligence tools or machine learning technologies. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)
Database:	Engineering Source

Description
Abstract:	Multi-class imbalanced medical data classification is one of the most critical challenges in contemporary machine learning, as it reflects the diverse nature of disease conditions. While considerable efforts have been made to address class imbalance, recent studies have increasingly recognized that class overlap poses an even greater challenge to classification performance than the imbalance itself. A promising future direction to address the class overlap issue is to explore the complex inter-class relationships present in real-world datasets. In this context, the Similarity Oversampling and Undersampling Preprocessing (SOUP) method, which combines oversampling and undersampling techniques, has been considered a potential baseline for handling multi-class imbalance and provide useful insights for addressing class overlap. In this study, we revisit the SOUP method and examine its strengths and limitations in handling class overlap. Based on this analysis, we experimentally examine two potential modifications to the SOUP method: integrating feature selection and removing the undersampling component. Experiments on ten multi-class imbalanced medical datasets show that both lead to improved classification performance under class overlap conditions, providing a meaningful direction for future research. [ABSTRACT FROM AUTHOR]
ISSN:	1819656X