Self-Adaptive Dragonfly Optimizer with Random Forest and Kernel Principal Component Analysis for Complex Optimization Problems

Optimization algorithms (OAs) frequently address complex industrial automation and system scheduling design problems. However, when handling high-dimensional complex problems, these algorithms often struggle to explore the entire solution space due to their reliance on extensive computational resources and significant time costs. Additionally, they are prone to becoming trapped in local optima when solving high-dimensional problems. To address these challenges, this work proposes a novel OA named Self-adaptive Dragonfly Optimizer with random Forest and kernel-Principal component analysis (SDOFP). The Self-adaptive Dragonfly Optimizer (SDO) is designed to dynamically adjust parameters for adaptive exploration of the decision space. Kernel-Principal Component Analysis (K-PCA) is a dimensionality reduction method, turning the high-dimensional search space into a lower-dimension one. It can effectively guide the population toward the global optimum. Random Forest (RF) is utilized as a surrogate model to balance training time and prediction accuracy, conserving computational resources and enhancing overall performance. The performance of SDOFP is evaluated by comparing it with several typical peers across eight high-dimensional benchmark functions. Further validation is conducted by applying SDOFP to a real-world flexible job shop scheduling problem, demonstrating its practical applicability.