Joint surrogate-assisted federated evolutionary feature selection algorithm

doi:10.1016/j.knosys.2026.116821

Joint surrogate-assisted federated evolutionary feature selection algorithm

2026-08-22

0

PRE

AI

Y

Ying Hu

张勇 (Yong‐Wei Zhang) *

DOI：10.1016/j.knosys.2026.116821

Original

Original request for help

Save

Abstract

En 中文

Evolutionary algorithms face the challenges of high computational cost and the “curse of dimensionality” in high-dimensional feature selection (FS) problems. These challenges are exacerbated when data representing the same learning task are distributed horizontally among multiple institutions that cannot share sensitive data. To address the high-dimensional FS problem with multiparticipation under privacy protection, a joint surrogate-assisted federated evolutionary feature selection algorithm (SaFEFS) is presented in this study. First, an XGBoost model-based joint filter FS approach is proposed to reduce the initial feature space while ensuring privacy. Subsequently, a surrogate-assisted federated evolutionary FS algorithm is designed, and several new strategies, including multiparticipant population initialization guided by feature importance, joint construction and management for surrogate models, and a joint individual updating strategy, are developed to enhance the performance of SaFEFS. Ultimately, SaFEFS is implemented on 12 test datasets and compared with several classical evolutionary FS algorithms. The experimental results demonstrate that SaFEFS can obtain feature subsets with good classification performance while preserving data privacy.

Keywords:

Feature selection

Evolutionary algorithm

Surrogate-assisted evolutionary algorithm

Privacy protection

Joint surrogate-assisted federated evolutionary feature selection algorithm

Abstract

Journal

Organization