Blobfish Optimization Algorithm: A Novel Met heuristic Method for High-Dimensional Global Optimization Problems

Abstract

Met heuristic optimization algorithms have been widely employed to solve complex nonlinear and high-dimensional optimization problems; however, many existing methods suffer from premature convergence, sensitivity to parameter settings, and scalability limitations. In this paper, a Blowfish Optimization Algorithm (BOA) is proposed as a sequential hybrid met heuristic framework that integrates elite-guided exploitation with centroid-based attraction, clustering-based diversity preservation, and stagnation-aware perturbation mechanisms in an iterative loop. The proposed algorithm is evaluated on a comprehensive benchmark suite comprising 14 unimodal, multimodal, ill-conditioned, and composite test functions across dimensionality ranging from 2 to 100. Experimental results demonstrate that BOA achieves reliable convergence across diverse problem landscapes, maintaining a 100% success rate for all benchmark functions up to 50 dimensions and preserving stable performance at higher dimensionality, with 90% success rate for Composition1 function at 100 dimensions and 70% success rate for Ackley function at 100 dimensions, reflecting the challenges of highly multimodal and composite landscapes in extreme dimensions. Convergence analysis reveals the algorithm's ability to escape local minima through its surprise attack mechanism. A detailed parameter sensitivity analysis confirms strong robustness to moderate parameter variations, with 100% success rate maintained across all 27 tested parameter combinations. Overall, the results suggest that BOA provides a robust and scalable hybrid optimization framework suitable for complex continuous optimization problems, offering a balanced trade-off between solution quality, robustness, and computational cost.