After an in-depth exploration of practical engineering problems, it has been discovered that the demand for solving expensive high-dimensional multi-objective optimization problems is gradually increasing. When traditional regression models are used to tackle such issues, cumulative error and computational complexity tend to surge significantly. To enhance the search efficiency of agent-assisted evolutionary algorithms and strike a balance between convergence and diversity in high-dimensional multi-objective problems, this paper introduces a Fuzzy Classification Preselection Strategy-based Agent-assisted Evolutionary Algorithm (FCPSEA). Initially, the population is initialized and evaluated, and two training sample sets are constructed using non-dominated relationships and congestion degree. Subsequently, the training samples and a double-archive operator guide the classifier to categorize more accurately. Finally, a model management strategy is proposed based on fuzzy classification preselection, which is set according to the predicted double-archive class labels and membership degrees. To validate the performance of the proposed algorithm, comparative experiments are conducted with classical algorithms in recent years on two groups of test problems encompassing various features. The experimental results demonstrate that the algorithm exhibits strong competitiveness in solving expensive high-dimensional multi-objective optimization problems.