MWMOTE-FRIS-INFFC: An Improved Majority Weighted Minority Oversampling Technique for Solving Noisy and Imbalanced Classification Datasets
Year: 2025
Type: article
Abstract: The utilization of high-noise unbalanced data poses significant challenges. In order to address this issue, an Enhanced Majority Weighted Minority Oversampling Technique is introduced, named MWMOTE-FRIS-INFFC, which specifically tackles noisy and imbalanced classification datasets. This approach leverages Euclidean distance for the allocation of sample weights, synthesizes, and incorporates novel samples into those of greater weights but belonging to minority classes, thereby addressing the data scarcity within smaller class clusters. Subsequently, the application of the fuzzy rough instance (FRIS) selection methodology eliminates a subset of synthetic minority samples exhibiting low cluster membership, effectively mitigating the overfitting tendency associated with synthetic oversampling of minority samples. Furthermore, the integration of the classification fusion iterative filter (INFFC) serves to mitigate synthetic noise issues, including both original and synthesized data noise. Following this, a series of experiments are designed to enhance the performance of six oversampling algorithms across eight datasets, utilizing the proposed MWMOTE-FRIS-INFFC algorithm.
Source: Applied Sciences
DOI: 10.3390/app15094670
Authors: Dong ZHANG, XIANG HUANG, GEN LI, SHENGJIE KONG, LIANG DONG
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: gold
Estimated APC paid: $2,896
OpenAlex ID: W4409729904