A novel interpretable machine learning and metaheuristic-based protocol to predict and optimize ciprofloxacin antibiotic adsorption with nano-adsorbent.

The existence of antibiotics in water sources poses substantial hazards to both the environment and public health. To effectively monitor and combat this problem, accurate predictive models are essential. This research focused on employing machine learning (ML) techniques to construct some models for analyzing the adsorption capacity of ciprofloxacin (CIP) antibiotic from contaminated water. The robustness of ten machine learning algorithms was evaluated using performance metrics such as the Coefficient of determination (R), Mean Square Error (MSE), Median Absolute Error (MedAE), Mean Absolute Error (MAE), Correlation coefficient (R), Nash-Sutcliffe Efficiency (NSE), Kling-Gupta Efficiency (KGE), and Root Mean Square Error (RMSE). The hyperparameters of the ML models were fine-tuned using the Bayesian optimization algorithm. The optimized models were comprehensively evaluated using feature importance analysis to quantify the relative significance of operational variables accurately. After a thorough assessment and comparison of various machine learning models, it was evident that the HistGradientBoosting (HGB) model outperformed others in terms of CIP adsorption performance. This was supported by their low MAE value of 0.1865 and high R value of 0.9999. The modeling projected the highest antibiotic adsorption (99.28%) under optimized conditions, including 10 mg/L of CIP, 357 mg/L of CuWO@TiO adsorbent, a contact time of 60 min at room temperature, and near neutral pH (7.5). The combination of advanced ML algorithms and nano adsorbents has great potential for addressing the problem of antibiotic pollution in water sources.