Feasibility of the Optimal Design of AI-Based Models Integrated with Ensemble Machine Learning Paradigms for Modeling the Yields of Light Olefins in Crude-to-Chemical Conversions