A Novel Explainable Nature-inspired Metaheuristic: Jaguar Algorithm with Precision Hunting Behavior
Ching-Hsuan Wu, Jyun–Yi Shen, Pei-Shin Huang, Cheng-Yen Hua, Yu-Chi Jiang, Shu–Yu Kuo, Yao–Hsin Chou · 2022 IEEE International Conference on Systems, Man, and Cybernetics (SMC) · 2022
Metaheuristics are crucial for solving complex optimization problems effectively in many fields. With the recent increased interest in explainable artificial intelligence (XAI), the issue of how metaheuristics interact with XAI has attracted much attention. Most metaheuristics have stochastic search processes, and the jaguar algorithm (JA) is a unique algorithm that has exact search paths. JA shows potential abilities both in exploration and exploitation, but still faces limitations. Therefore, this study proposes a new metaheuristic based on JA and invents precision hunting behavior (PH-JA), which inherits the concept of JA and significantly strengthens its search efficiency. PH-JA improves the solution quality by adaptively detecting the current environment’s trends during movement and precisely hunting prey. Then, PH-JA makes the best of the historical information to reduce the computational cost. PH-JA is an explainable metaheuristic and has systematic operational procedures in that each movement is performed according to the present situation rather than random elements. Therefore, the path to a solution is interpretable and can return the same result with the same input. The experimental results demonstrated the superiority of PH-JA, which is better than other classical and state-of-the-art metaheuristics in terms of solution quality and evaluation costs.