Guiding experiment with Machine Learning: A case study of biochar adsorption of Ciprofloxacin

[Display omitted] •Extracting pertinent data from 50 relevant published articles.•Developing an automated hyperparameter search machine learning model.•The predictions of the four machine learning models are compared.•Utilizing SHAP to scrutinize the model's predictive outcomes.•Suggested values for 20 features are provided to guide the experiments. This study employs an integrated approach, using K-FOLD cross-validation, Optuna automated hyperparameter search, machine learning, and SHapley Additive exPlanations (SHAP) analysis, to examine factors influencing biochar's capacity to adsorb the antibiotic ciprofloxacin (CIP), a significant contributor to global water pollution. We ensured our model's reliability and robustness using K-FOLD, then optimized its predictive accuracy with Optuna. We accurately predicted biochar's adsorption capacity using 20 distinct features. Utilizing SHAP, we unraveled each feature's intricate interactions and contributions towards the adsorption capacity, subsequently determining the optimal value range for each feature that significantly improves biochar's efficiency in CIP adsorption. Our research promotes biochar's application in environmental governance, particularly in treating antibiotic wastewater. By providing a profound understanding and optimizing adsorption characteristics, our findings hold substantial reference value for improving the efficiency of biochar's application in environmental protection and guiding associated experimental design and preparation of biochar.