High-dimensional multi-objective feature selection with niche-based binary differential evolution

•The model for the multi-objective feature selection problem has been established.•A niche-based binary differential evolution algorithm is proposed.•The MONBDE algorithm outperforms the comparison algorithms in terms of performance. Feature selection is a critical step in machine learning and data...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:Expert systems with applications Jg. 298; S. 129478
Hauptverfasser: Yue, Xuezhi, Zuo, Xiang, Ling, Pengfei, Xiong, Chao, Peng, Hu, Zeng, Yuan
Format: Journal Article
Sprache:Englisch
Veröffentlicht: Elsevier Ltd 01.03.2026
Schlagworte:
Differential evolution
Feature selection
Multi-objective optimization
Redundant solution repair
Differential evolution
Feature selection
Multi-objective optimization
Redundant solution repair
ISSN:0957-4174
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:•The model for the multi-objective feature selection problem has been established.•A niche-based binary differential evolution algorithm is proposed.•The MONBDE algorithm outperforms the comparison algorithms in terms of performance. Feature selection is a critical step in machine learning and data mining, aiming to identify the most relevant features from a dataset to improve model performance while reducing computational costs. In high-dimensional data, as the dimensionality of data increases rapidly, feature selection faces an enormous search space, limiting the efficiency and effectiveness of traditional methods. To address these challenges, multi-objective optimization algorithms have emerged as a promising strategy for feature selection due to their ability to optimize multiple conflicting objectives simultaneously. We propose a niche-based binary differential evolution algorithm (MONBDE) for high-dimensional multi-objective feature selection. MONBDE enhances feature selection performance through several mechanisms: a niche-based binary differential evolution operator, redundant solution repair mechanism and an environmental selection strategy. In experiments, the proposed algorithm was compared with five advanced multi-objective optimization algorithms and tested on 15 benchmark datasets using three common metrics. Experimental results show that the MONBDE algorithm outperforms comparative algorithms in terms of classification accuracy and feature subset size across most datasets. The proposed strategy effectively eliminates redundant and irrelevant solutions in feature selection, leading to a significant improvement in model classification performance.
ISSN:0957-4174
DOI:10.1016/j.eswa.2025.129478