Application of machine learning methods for phenols degradation by Fe-based catalytic Fenton-like process.

Long Short-Term Memory most accurately predicts phenols removal efficiency in the curve-group of Fe-based catalytic Fenton-like systems, reaching a pooled out-of-fold R² of 0.883.

• Why it matters: Interactions among catalyst composition, structure, water matrix, and operating conditions make removal performance hard to predict and mechanisms difficult to compare across reaction systems.
• What they did: A literature-built dataset covering phenol properties, Fe-based catalyst composition/structure, aqueous matrix, and reaction conditions was used to train machine learning and deep learning models, evaluated with grouped nested cross validation.
• The result: Model accuracy depended on validation grouping—Random Forest reached 0.708 (system-group) and 0.681 (catalyst-group), while Extreme Gradient Boosting reached 0.589 (paper-group)—and SHAP/PFI/VIF highlighted reaction time and pH as most stable factors.

AI summary of the abstract · may contain errors · not medical advice · read the paper.