Abstract
As a new algorithm in recent years, the Kernel Search Optimization (KSO) algorithm employs kernel mapping to tackle optimization challenges, but this approach can lead to accuracy loss when transforming the objective function from lower to higher dimensions. Moreover, the mapping approximation might fail to capture the true optimal solution, particularly in complex, high-dimensional scenarios. To address these limitations, we introduce an enhanced version: the Chaotic map, Adaptive t-distribution Mutation, and Sand Cat Behavior with Spiral Search (CSTKSO) algorithm. This enhanced approach utilizes chaotic mapping for population initialization, lowering the likelihood of premature convergence. It also features Adaptive t-distribution mutation, which perturbs solution positions and dynamically adjusts the degrees of freedom parameter based on iteration progress, balancing global exploration in early stages with local exploitation later. Additionally, sand cat behavior-inspired mechanisms, including random angle selection, enable the algorithm to obtain the best results in a wide range of search optimization. We evaluated CSTKSO against established algorithms using 50 benchmark functions from the IEEE CEC for real-parameter optimization. The findings highlight the exceptional performance of CSTKSO in comparison to the original KSO and other algorithms. Furthermore, when applied to a practical economic emission scheduling problem, CSTKSO outperformed other competing algorithms, demonstrating its effectiveness in real-world applications. This enhanced algorithm addresses the shortcomings of traditional KSO while maintaining its core strengths, offering a more robust and efficient optimization tool for complex problems.