Abstract
<title>Abstract</title> <p>Expensive high-dimensional multi-objective optimization problems pose a model-state mismatch challenge to surrogate-assisted evolutionary algorithms. To address this issue, this paper proposes HDFC-ASS, a classification-regression cooperative fuzzy surrogate-assisted evolutionary algorithm. The proposed method first designs a variable-correlation-guided random-subspace incremental Kriging strategy to acquire informative samples in high-dimensional decision spaces. Then, a fuzzy classifier-assisted local exploitation mechanism with a balanced-accuracy-based one-way transition is developed to prevent premature use of unreliable classification boundaries. Finally, a convergence-diversity-uncertainty cooperative model-management criterion is proposed to allocate expensive evaluations among global exploration, boundary correction and high-membership exploitation. The proposed framework is designed for expensive high-dimensional multi-objective optimization under a strictly limited evaluation budget.</p>