A novel interpretable machine learning and metaheuristic-based protocol to predict and optimize ciprofloxacin...

Area of Science:

• Environmental Science
• Water Chemistry
• Computational Chemistry

Background:

• Antibiotic contamination in water poses significant environmental and public health risks.
• Accurate predictive models are crucial for monitoring and mitigating antibiotic pollution.

Purpose of the Study:

• To develop and evaluate machine learning (ML) models for predicting the adsorption capacity of ciprofloxacin (CIP) from contaminated water.
• To identify the optimal ML model and operational conditions for efficient CIP removal.

Main Methods:

• Evaluated ten ML algorithms using metrics like R², MSE, MAE, and RMSE.
• Fine-tuned ML model hyperparameters using Bayesian optimization.
• Assessed model performance and feature importance for operational variables.

Main Results:

• The HistGradientBoosting (HGB) model demonstrated superior performance with MAE of 0.1865 and R² of 0.9999.
• Projected 99.28% CIP adsorption under optimized conditions: 10 mg/L CIP, 357 mg/L CuWO₄@TiO₂ adsorbent, 60 min contact time, room temperature, and pH 7.5.
• Feature importance analysis confirmed the significance of operational variables.

Conclusions:

• Advanced ML algorithms, particularly HGB, are effective for modeling antibiotic adsorption.
• Optimized conditions using nano-adsorbents show high potential for removing ciprofloxacin from water.
• This approach offers a promising strategy to combat antibiotic pollution in water sources.