Dynamic Enzyme Action Optimizer (dEAO): An Adaptive Bio-Inspired Approach for Classical Benchmark Functions
Serdar Ekinci, Davut İzci, Gökhan Yüksek, Rıdvan Fırat Çınar, Mostafa Rashdan, Mohammad Salman · 2025
This study introduces the dynamic enzyme action optimizer (dEAO), a novel bio-inspired metaheuristic that builds upon and significantly enhances the original enzyme action optimizer (EAO). The core idea of dEAO draws inspiration from enzymatic catalysis, modeling solution updates based on enzyme-substrate dynamics. While EAO provides a strong exploratory foundation, its performance is limited by fixed control parameters and a lack of local refinement. To overcome these drawbacks, dEAO incorporates three key modifications: (i) a sinusoidally adaptive factor that dynamically balances exploration and exploitation throughout the search process, (ii) a randomized enzyme concentration to introduce controlled stochasticity and avoid premature convergence, and (iii) a lightweight local search phase to finetune solutions in later iterations. These enhancements allow dEAO to adapt its search behavior based on problem complexity and iteration context. The algorithm is evaluated on four widely used benchmark functions (Rosenbrock, Step, Schwefel, and Penalized) representing diverse and challenging optimization landscapes. Experimental results demonstrate that dEAO consistently outperforms its predecessor. The comparative analysis validates the proposed modifications, showing that dEAO achieves lower mean errors and superior best-case results across all test functions.