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

Area of Science:

• Optimization Algorithms
• Computational Intelligence
• Metaheuristics

Background:

• Metaheuristic algorithms are crucial for solving complex optimization problems.
• Existing algorithms often face challenges with exploration-exploitation balance and maintaining population diversity.
• Physics-inspired algorithms offer novel mechanisms for search optimization.

Purpose of the Study:

• To propose Gyro-Top Optimization (GTO), a new metaheuristic algorithm inspired by gyroscope dynamics.
• To evaluate GTO's performance on standard benchmark functions and compare it with existing algorithms.
• To assess GTO's applicability to real-world problems like feature selection and engineering optimization.

Main Methods:

• GTO employs a staged search framework incorporating precession and nutation dynamics.
• Lévy-flight perturbation and elite-gyro collision enhance global exploration.
• Precession-guided search and nutation-based perturbation facilitate local exploitation.
• A dynamic elite pool using cosine-similarity k-means clustering maintains population diversity.

Main Results:

• GTO achieved first rank on CEC2017 and CEC2020 benchmark functions across dimensions.
• GTO ranked within the top four on CEC2014 benchmark functions.
• Statistical tests confirmed GTO's significant advantages over many algorithms, showing competitiveness in high-dimensional cases.
• GTO demonstrated competitive performance in high-dimensional biomedical feature selection and engineering optimization problems.

Conclusions:

• GTO is a robust and effective metaheuristic optimization algorithm.
• The algorithm shows strong performance on numerical, combinatorial, and constrained engineering problems.
• GTO's physics-inspired approach offers a promising direction for developing advanced optimization techniques.