Evolutionary polar fox optimization algorithm: an enhanced metaheuristic for feature selection

doi:10.1007/s10586-026-06402-4

Evolutionary polar fox optimization algorithm: an enhanced metaheuristic for feature selection

2026-08-06

0

PRE

AI

DOI：10.1007/s10586-026-06402-4

Original

Original request for help

Save

Abstract

En 中文

Feature selection can present challenges such as premature convergence and difficulty balancing classification performance with subset compactness. This study addresses these issues by proposing an Evolutionary Polar Fox Algorithm (EPFA), which integrates genetic crossover and mutation into the experience stage and introduces particle swarm velocity guidance into the leadership stage of the original Polar Fox Algorithm. This hybrid mechanism improves both population diversity and convergence stability, all the while maintaining the lightweight search structure of the Polar Fox Algorithm (PFA). The EPFA was evaluated using the CEC2021 and CEC2020 benchmark suites, as well as ten publicly available UCI datasets for feature selection. Experimental results demonstrate that the EPFA achieves faster convergence, higher solution accuracy and more compact feature subsets than several representative metaheuristic algorithms. Furthermore, ablation experiments verify the contributions of the genetic algorithm (GA) and particle swarm optimisation (PSO) components. These results confirm the effectiveness and general applicability of the proposed method.

Keywords:

Feature selection

Evolutionary polar fox algorithm

Genetic algorithm

Particle swarm optimization

Numerical optimization

Evolutionary polar fox optimization algorithm: an enhanced metaheuristic for feature selection

Abstract

Journal

Organization