Gyro-top optimization: a physics-inspired metaheuristic for engineering optimization and feature selection

This paper proposes Gyro-Top Optimization (GTO), a physics-inspired metaheuristic algorithm motivated by the precession and nutation behaviors of a spinning gyroscope. GTO translates gyroscopic dynamics into a staged search framework, where Lévy-flight perturbation and elite-gyro collision enhance global exploration, while precession-guided search and nutation-based perturbation support local exploitation. To maintain diversity and reduce excessive elite concentration, a dynamic elite pool based on cosine-similarity k-means clustering is further incorporated. The performance of GTO is evaluated on 69 benchmark functions from the IEEE CEC2014, CEC2017, and CEC2020 test suites under 10D, 30D, 50D, and 100D settings, with comparisons against 15 representative metaheuristic algorithms. The Friedman ranking results show that GTO ranks first on all tested dimensions of CEC2017 and CEC2020, and remains within the top four on all CEC2014 dimensions. Wilcoxon signed-rank tests further confirm that GTO achieves statistically significant advantages over many compared algorithms while remaining comparable to several strong competitors in some high-dimensional cases. Additional parameter sensitivity and ablation studies verify the robustness of the default configuration and the contribution of key components, especially precession, population reduction, and cosine-based elite maintenance. GTO is also applied to high-dimensional biomedical feature selection and 20 selected constrained engineering problems from the Enoppy library. The application results indicate that GTO can achieve competitive feature-selection performance and the best Friedman average rank on the selected engineering benchmark set. Overall, the results demonstrate that GTO is a competitive and practically applicable optimizer for numerical, combinatorial, and constrained engineering optimization problems. The source code of GTO is available at https://github.com/T3t5uy4/Gyro-Top-Optimization.