Hybridizing Artificial Electric Field Algorithm with Lévy Flight and Chaotic Dynamics for Enhanced Optimisation Performance

The Artificial Electric Field Algorithm (AEFA) suffers from premature convergence and local minima entrapment, limiting its effectiveness in complex optimisation scenarios. To address these limitations, we propose the Lévy-Flight and Chaos-based Artificial Electric Field Algorithm (LCAEFA), which synergistically combines Lévy flight distribution for enhanced global exploration and chaotic dynamics for improved search diversity. The Lévy flight mechanism enables particles to perform strategic long-distance jumps guided by power-law distributions, while ten distinct chaotic maps introduce controlled perturbations that prevent stagnation in local optima. This dual enhancement creates an optimal balance between exploration and exploitation phases throughout the optimisation process. LCAEFA is rigorously evaluated on six benchmark functions spanning unimodal, multimodal, and fixed-dimensional categories, demonstrating superior convergence rates and solution quality compared to the original AEFA. Furthermore, we validate LCAEFA’s practical applicability by employing it as a trainer for Multilayer Perceptron (MLP) neural networks across five MLP training benchmarks: two real-world medical classification datasets (breast cancer, heart disease), one synthetic classification benchmark (XOR), and two synthetic function approximation tasks (sigmoid, cosine). Comparative analysis against ten state-of-the-art heuristic algorithms reveals that LCAEFA achieves up to 100% classification accuracy on the XOR benchmark and 88% on the breast cancer dataset. Statistical validation through Wilcoxon signed-rank tests confirms the significance of performance improvements. The integration of Lévy flight and chaotic dynamics successfully transforms AEFA into a robust optimiser capable of handling diverse optimisation challenges with enhanced convergence characteristics and superior solution quality.