Md. Rubel, Cries Avian, M.M. Harussani, Eric Kolor, Sasipa Boonyubol, Koichi Mikami, Muhammad Aziz, Jeffrey S. Cross
Biodiesel production from waste cooking oil (WCO) is fundamentally constrained by complex, nonlinear interactions among transesterification parameters, which conventional linear optimization fails to capture. This study addresses this knowledge gap by developing an automated ensemble machine learning (AutoML) framework within the PyCaret environment. Utilizing 49 in-house experimental data points derived from a previously established two-step acid-base catalysis process, 25 machine learning (ML) regression algorithms were systematically benchmarked. The top 15 models were optimized via Optuna's Tree-structured Parzen Estimator (TPE) Bayesian method to construct bagging, boosting, and stacking ensemble strategies. The boosting ensemble model served as a process interaction mapping framework, achieving superior performance under 3-fold cross-validation (R² = 0.987, MAE = 0.599, and RMSE = 0.836), significantly outperforming individual learners as well as bagging and stacking ensemble architectures. Shapley Additive exPlanations (SHAP) analysis served as a sophisticated tool for operational boundary identification, deconstructing model sensitivity regarding reaction temperature and time. Furthermore, external validation using 123 literature-derived points demonstrated significant predictive capability (R² = 0.973), proving the boosting ensemble's promising generalizability across related WCO-based transesterification datasets. This framework establishes an operational space mapping protocol, identifying biodiesel yield and optimal transesterification conditions through data-driven modeling. © 2026 The Authors
Department of Transdisciplinary Science and Engineering, School of Environment and Society, Institute of Science Tokyo, 2-12-1 Ookayama, Meguro-ku, Tokyo, 152-8550, Japan; Department of Electrical Engineering, Universitas Brawijaya Malang, Jawa Timur, 65145, Indonesia; Institute of Industrial Science, The University of Tokyo, 4-6-1 Komaba, Meguro-ku, Tokyo, 153-8505, Japan
