TPFS: A three-phase heuristic feature selection algorithm for large-scale sample datasets

doi:10.1016/j.patcog.2025.112084

TPFS: A three-phase heuristic feature selection algorithm for large-scale sample datasets

2025-07-07

0

PRE

AI

DOI：10.1016/j.patcog.2025.112084

Original

Original request for help

Save

Abstract

En 中文

• We propose a novel three‑phase heuristic feature selection algorithm, termed TPFS, designed to tackle the twin challenges of large‑scale data compression and the identification of an optimal feature subset. • TPFS initially selectively retains representative samples to dramatically lower memory demands while preserving essential data characteristics. During hypergraph initialization, it enhances the evolutionary algorithm by solving vertex cover problem, yielding more stable population and faster convergence. Finally, in the whale optimization phase, TPFS utilizes a dual-mode operation and integrates multiple classifiers (CART, SVM, and KNN) to bolster generalization and robustness in feature selection. • Extensive experiments demonstrate that TPFS outperforms seven state of-the-art algorithms, especially in terms of scalability and accuracy when handling large-scale sample datasets.

TPFS: A three-phase heuristic feature selection algorithm for large-scale sample datasets

Abstract

Journal

Organization