An Adaptive Grouping-Based Hybrid PSO-GA Algorithm for Enhanced Global Optimization
This paper proposes an adaptive grouping-based hybrid PSO-GA algorithm to address PSO's premature convergence and GA's slow convergence. The algorithm dynamically partitions the population into elite/regular subgroups, applying PSO for local exploitation and GA for global exploration. Key innovations include adaptive grouping, dynamic weight adjustment, bidirectional elite migration, and stagnation-triggered multi-phase optimization. Experiments on 7 benchmarks show superior convergence accuracy, speed, and stability over standalone PSO/GA (especially for multimodal problems). Further validation on highdimensional $(\mathbf{D}=\mathbf{5 0} / \mathbf{1 0 0})$ CEC 2017 benchmarks, constrained engineering designs, and sensitivity analysis confirms its scalability, robustness, and practical applicability.