Automated machine learning-based models for predicting and evaluating antibiotic removal in constructed wetlands

Area of Science:

• Environmental Engineering
• Water Treatment Technologies
• Machine Learning Applications

Background:

• Constructed wetlands (CWs) are vital for removing antibiotics from wastewater.
• Optimizing CW operation is crucial for effective antibiotic removal.
• Current modeling approaches struggle to capture the complex biochemical processes involved.

Purpose of the Study:

• To develop robust machine learning models for predicting antibiotic removal in CWs.
• To identify key operational variables influencing antibiotic removal performance.
• To provide a framework for optimizing CW processes using data-driven insights.

Main Methods:

• Two automated machine learning (AutoML) models were developed and trained on CW operational data.
• Model performance was evaluated using metrics like mean absolute error and coefficient of determination.
• Explainable AI techniques, including variable importance and Shapley additive explanations, were employed.

Main Results:

• AutoML models demonstrated strong predictive performance for antibiotic removal across different training dataset sizes.
• Substrate type was identified as a more significant factor than influent wastewater quality or plant type.
• The models successfully predicted antibiotic removal without requiring human intervention in the operational adjustments.

Conclusions:

• Automated machine learning offers a powerful tool for predicting and optimizing antibiotic removal in constructed wetlands.
• Understanding the influence of operational variables, particularly substrate type, is key to enhancing CW efficiency.
• This study provides a valuable reference for data-driven operational adjustments in CW systems for improved antibiotic removal.