Federated feature selection algorithm for multi-party high-dimensional data under privacy protection
JIN Xin
HU Ying
LI Peng
ZHENG Ming
Abstract：High-dimensional feature selection faces challenges such as the"curse of dimensionality"and high computational costs.Due to privacy protection constraints,a large amount of high-dimensional data may be distributed and stored across different institutions(referred to as participants)and cannot be shared,further complicating the joint feature selection of multi-party high-dimensional data.In light of this,this paper proposes a surrogate-joint-assisted federated evolutionary feature selection algorithm to ad-dress the issue of high-dimensional feature selection involving multiple participants under privacy protec-tion.A framework for a surrogate-assisted federated evolutionary feature selection algorithm is designed,and based on this framework,strategies for joint construction and management of surrogate models,joint evaluation based on surrogate models,and joint updating of individuals are provided.Finally,the proposed algorithm is applied to 10 test datasets and compared with 3 typical wrapper-based evolutionary feature se-lection algorithms.The results show that the proposed algorithm not only ensures the classification per-formance of the algorithm while fully protecting the data privacy of the participants but also significantly improves the algorithm's runtime.
Keywords：feature selectionevolutionary algorithmsurrogate-assistedprivacy protection
Publication Date：2025-08-01
Online Publishing Date：2026-08-28(First online date of this platform, not the publication date of the document)
Pages：10( 497-506 )