.png){kind=icon}
Return
A high-dimensional feature selection method based on multiobjective evolutionary algorithm with multilayer search strategy for imbalanced multiclass data
.png){kind=icon}
PRE
AI
DOI：10.1016/j.eswa.2026.133004.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Evolutionary algorithms (EAs) have emerged as an effective solution for feature selection (FS). However, high-dimensional datasets with class imbalance that are common in the real world continue to pose a significant challenge for EA-based FS methods. The imbalance issue is particularly complex in multiclass classification scenarios, where efficient FS methods are still scarce. To address these issues, a multiobjective evolutionary feature selection algorithm based on a multilayer search strategy (MLS-MOEA) is proposed in this paper. The proposed method integrates structure-aware feature grouping with a multilayer search mechanism to improve search efficiency while maintaining the quality of selected feature subsets. Specifically, feature clustering is employed to reduce the search space, and a cooperative search strategy is applied at both the cluster and feature levels. In addition, a label-aware evolutionary mechanism is incorporated to better preserve features that are relevant to minority classes. The effectiveness of the proposed approach is evaluated on multiple high-dimensional imbalanced benchmark datasets and a real-world industrial dataset. Experimental results show that MLS-MOEA achieves competitive or superior performance compared with seven state-of-the-art evolutionary feature selection methods in terms of multiobjective metrics and classification performance on most datasets. These results suggest that integrating clustering and multilayer search within a multiobjective evolutionary framework can be a useful strategy for feature selection in high-dimensional imbalanced multiclass scenarios.
Keywords:
High-dimensional feature selection
Multiobjective evolutionary algorithm
Multilayer search
Imbalanced data
Multiclass classification
Journal
IF:
7.5
Papers:
3.0W
Citations:
10.2W
Organization
Cited Papers
No cited papers available