Superb Fairy-wren Optimization Algorithm: a novel metaheuristic algorithm for solving feature selection problems

This paper presents the Superb Fairy-wren Optimization Algorithm (SFOA), a novel meta-heuristic optimization algorithm based on swarm intelligence. The SFOA is proposed for numerical optimization, engineering design, and high-dimensional feature selection classification, drawing inspiration from the life habits of the Superb fairy-wren. The algorithm incorporates the development of young birds in Superb fairy-wren colonies, the behavior of the birds to feed their young after breeding, and their tactics to avoid predators. By simulating three natural behaviors of the Superb fairy-wren corresponding to the stages of young bird growth, breeding and feeding, and avoiding natural enemies, a mathematical model is established. Specific judgment conditions are employed to enable the algorithm to efficiently and accurately complete optimization tasks across different areas of the search space. To evaluate the optimization effect of SFOA, a comparison is made with several classical optimization algorithms and novel optimization algorithms using qualitative and quantitative methods, employing benchmark test functions from IEEE CEC2017 and IEEE CEC2022. The comparisons demonstrate the high efficiency and robustness of SFOA. Furthermore, the proposed algorithm is applied to fourteen CEC constrained engineering design problems, five constrained engineering design problems and a high-dimensional feature selection problem with packaging method, and the practicability of the proposed algorithm is verified. The experimental results further validate the high practicality and efficiency of SFOA in solving real-world application problems.