A state-adaptive booby optimization algorithm for engineering design and medical data applications

Balancing global exploration and local exploitation remains a central challenge in metaheuristic optimization, particularly for high-dimensional, nonlinear, and constrained problems encountered in engineering design and medical data analysis. This paper proposes the Booby Optimization Algorithm (BOA), a state-adaptive population-based metaheuristic inspired by avian dive-foraging behavior but formulated entirely through mathematical and computational mechanisms. BOA employs an adaptive state variable to regulate step magnitude and dynamically control transitions between global exploratory search and local exploitative refinement, augmented by nonlinear motion dynamics, stochastic perturbations, and a recovery strategy for diversity preservation. The algorithm is extensively evaluated on CEC benchmark functions with dimensionalities up to 100, four classical constrained engineering design problems, and feature selection and classification tasks on 14 real-world medical datasets using BOA-based hybrid models. Experimental results demonstrate that BOA consistently outperforms several state-of-the-art metaheuristics in terms of convergence speed, solution accuracy, and robustness, achieving near-optimal or best-known solutions with significantly reduced mean error and variance. In medical classification tasks, BOA-based feature selection attains a mean accuracy of 96.20% ± 1.05, alongside high sensitivity and specificity while effectively reducing feature dimensionality. These improvements are supported by rigorous statistical validation using Friedman, Nemenyi, and Wilcoxon tests (p < 0.001). Overall, the results establish BOA as an efficient and robust adaptive optimization framework suitable for complex engineering optimization and medical decision-support applications.