A Novel Hybrid Resampling Based on Relative Class‐Impurity for Class Imbalanced Problems.
Saved in:
| Title: | A Novel Hybrid Resampling Based on Relative Class‐Impurity for Class Imbalanced Problems. |
|---|---|
| Authors: | Park, You-Jin1 (AUTHOR) yjpark@mail.ntut.edu.tw, Cheng, Ke-Yong2 (AUTHOR), Pan, Rong3 (AUTHOR), Montgomery, Douglas C.3 (AUTHOR), Dang, Quang-Vinh4 (AUTHOR), Lee, Juseong4 (AUTHOR), Mohanta, Kallol (AUTHOR) kmohanta@wiley.com |
| Source: | International Journal of Intelligent Systems. 7/27/2026, Vol. 2026, p1-16. 16p. |
| Subjects: | Resampling (Statistics), Classification algorithms, Data reduction, Data augmentation |
| Abstract: | Classification is widely used to categorize instances and predict their membership of new instances in various learning applications. However, class imbalance is a common issue in many real‐world situations, where class distributions are not uniform. As traditional classification algorithms have been designed to handle balanced classification problems, class imbalance is one of the most significant challenges in modern learning tasks, resulting in degrading classification performance and increasing learning difficulty. Additionally, class overlap occurs when instances from multiple heterogeneous classes share a common feature space. Both class imbalance and class overlap problems are significant challenges in many practical learning tasks, since they may adversely affect the learning performances as well as increase the learning difficulty. Thus, in this study, to enhance the classification performance of imbalanced data with class overlap, a new hybrid resampling method based on relative class impurity, called CIHR, is proposed. In particular, the relative class impurity measure is considered an indicator of class overlap degree for adaptively applying oversampling and undersampling. The comprehensive experimental analysis considering several class imbalanced datasets and classification models demonstrates that the proposed method improves the classification performance as well as effectively mitigates the class imbalance and overlap problems. [ABSTRACT FROM AUTHOR] |
| Copyright of International Journal of Intelligent Systems is the property of Wiley-Blackwell 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 |
|
Full text is not displayed to guests.
Login for full access.
|
|
| Abstract: | Classification is widely used to categorize instances and predict their membership of new instances in various learning applications. However, class imbalance is a common issue in many real‐world situations, where class distributions are not uniform. As traditional classification algorithms have been designed to handle balanced classification problems, class imbalance is one of the most significant challenges in modern learning tasks, resulting in degrading classification performance and increasing learning difficulty. Additionally, class overlap occurs when instances from multiple heterogeneous classes share a common feature space. Both class imbalance and class overlap problems are significant challenges in many practical learning tasks, since they may adversely affect the learning performances as well as increase the learning difficulty. Thus, in this study, to enhance the classification performance of imbalanced data with class overlap, a new hybrid resampling method based on relative class impurity, called CIHR, is proposed. In particular, the relative class impurity measure is considered an indicator of class overlap degree for adaptively applying oversampling and undersampling. The comprehensive experimental analysis considering several class imbalanced datasets and classification models demonstrates that the proposed method improves the classification performance as well as effectively mitigates the class imbalance and overlap problems. [ABSTRACT FROM AUTHOR] |
|---|---|
| ISSN: | 08848173 |
| DOI: | 10.1155/int/5593646 |
