Feature selection using importance-based two-stage multi-modal multiobjective particle swarm optimization

Feature selection is aimed at reducing the dimensionality of datasets while maintaining or enhancing classification accuracy. Recent studies have increasingly approached feature selection through multi-modal, multi-objective optimization. However, for high-dimensional datasets, the large decision sp...

Full description

Saved in:
Bibliographic Details
Published in:Cluster computing Vol. 28; no. 2; p. 115
Main Authors: Ling, Qinghua, Liu, Wenkai, Han, Fei, Shi, Jinlong, Hussein, Ali Aweis, Sayway, Ben Sanvee
Format: Journal Article
Language:English
Published: New York Springer US 01.04.2025
Springer Nature B.V
Subjects:
Algorithms
Classification
Computer Communication Networks
Computer Science
Correlation coefficients
Datasets
Feature selection
Multiple objective analysis
Mutation
Operating Systems
Optimization algorithms
Optimization techniques
Pareto optimum
Particle swarm optimization
Processor Architectures
Speciation
Velocity
Feature importance
Feature selection
Multimodal multiobjective optimization
Particle swarm optimization (PSO)
Niching method
ISSN:1386-7857, 1573-7543
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Feature selection is aimed at reducing the dimensionality of datasets while maintaining or enhancing classification accuracy. Recent studies have increasingly approached feature selection through multi-modal, multi-objective optimization. However, for high-dimensional datasets, the large decision space limits the ability of traditional multi-modal, multi-objective optimization algorithms to effectively eliminate redundant features. To address this challenge, a two-stage multi-modal, multi-objective particle swarm optimization algorithm incorporating feature importance for feature selection is proposed. In the first stage, feature importance is evaluated by integrating spearman’s rank correlation coefficient and the maximal information coefficient, which facilitates the elimination of redundant and weakly correlated features, thus reducing the search space. In the second stage, the limitations of speciation-based niching algorithms in escaping local optima are addressed by introducing a mutation mechanism based on feature importance, which is applied to the optimal particles within each niche, enabling dominated particles to escape local optima. Experimental results demonstrate that our proposed method identifies lower-dimensional, superior equivalent feature subsets across ten datasets without compromising classification accuracy.
Bibliography:ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 14
ISSN:1386-7857
1573-7543
DOI:10.1007/s10586-024-04807-7