Group-driven Remora Optimization Algorithm with multiple search and regeneration strategies

We have introduced group-driven strategies into the conventional remora optimization algorithm (ROA) and amalgamated regeneration strategies to further enhance the generalization ability. This proposed algorithm were termed the Group-Driven Remora Optimization Algorithm (GROA). Firstly, the “Group dividing” divides the search individuals into different groups based on Euclidean distance, further subdividing individuals within each group with varying functionalities of search, using multiple search strategies. That provides different exploration scales and candidate solution diversity to the algorithm. Whereafter, if too many individuals are gathered together, “Random regeneration” randomly regenerates individuals. Finally, the offspring selection operation was produced for more efficient offspring evolution. To assess the effectiveness of the proposed GROA, the ablation experiments on the proposal and its derived variants, along with comparative experiments among GROA and the other five novelty and classical evolutionary algorithms (EAs) are conducted on CEC2017 test suits. Also, comparison experiments on five engineering design problems are implemented for appraising the performance in real-world optimization problems. Finally, we discuss the improvement and deficiency of the new algorithm, GROA. The experimental and statistical results demonstrate that our approach significantly improves the optimization accuracy and search efficiency of the conventional algorithm in most experimental problems.