Abstract
Particle Swarm Optimization (PSO) is a widely used algorithm for continuous optimization; however, its global search performance is often constrained by anisotropy in particle velocities. To mitigate this limitation without heavy computational overhead, this paper presents a generalized n-period hybrid PSO that alternates between a random linked PSO phase for global exploration and a random unlinked PSO phase for local exploitation. By incorporating swarm entropy analysis alongside success rates and best function values, we evaluate global search behavior and optimization performance across separable and non-separable benchmark functions in 10 and 20 dimensions. Numerical experiments show that the proposed n-period hybrid PSO exhibits higher swarm entropy during global exploration, suppresses velocity anisotropy, and outperforms standard PSO, original hybrid PSO, and the linearly decreasing inertia weight method (LDIWM). Notably, periodic alternation between global and local search phases significantly enhances search performance, with the 4-period hybrid PSO variant achieving the highest average success rate and the lowest best function values on the largest number of benchmark functions. Overall, the proposed framework provides a simple yet effective strategy for balancing exploration and exploitation in complex continuous optimization problems.