Distribution-Guided Hybridization of JADE and Nelder-Mead for Efficient Global Optimization under Evaluation Constraints

In real-world optimization problems, obtaining high-quality solutions with a limited number of function evaluations is essential. In this study, we propose a hybrid optimization method that dynamically switches between JADE and the Nelder-Mead method based on the distribution of individuals. The switching mechanism is designed to use JADE for global exploration, while applying the Nelder-Mead method when the distribution of individuals indicates stagnation, thereby enabling efficient local search. Experiments were conducted using CEC2017 benchmark functions under a significantly reduced evaluation budget compared to the standard maximum evaluation count of dimensions × 10000. As a result, the proposed method demonstrated superior performance over JADE, particularly under low-dimensional and low-evaluation conditions. This finding confrms that distribution-based switching is effective in hybrid evolutionary optimization under limited evaluation conditions.