Optimal selection of machine learning algorithms for ciprofloxacin prediction based on conventional water quality...

Area of Science:

• Environmental Science
• Analytical Chemistry
• Computational Science

Background:

• Antibiotics in aquatic environments pose risks to ecology and human health.
• Current antibiotic detection methods are often slow, costly, and labor-intensive.
• Rapid prediction methods for antibiotic concentrations are needed.

Purpose of the Study:

• To explore the feasibility of predicting ciprofloxacin (CFX) concentrations using conventional water quality indicators.
• To evaluate the effectiveness of machine learning algorithms for predicting antibiotic levels in aquatic environments.
• To identify the optimal machine learning model for accurate CFX concentration prediction.

Main Methods:

• Utilized three machine learning algorithms and two parameter optimization algorithms.
• Employed six input variables: COD, NH4+-N, DO, WT, TN, and pH.
• Evaluated model performance using R², NSE, RMSE, and MAPE metrics.

Main Results:

• The generalized regression neural network (GRNN) optimized by particle swarm optimization (PSO) demonstrated superior prediction accuracy.
• The best model achieved R²=0.936, NSE=0.915, RMSE=3.150 ng/L, and MAPE=30.909%.
• The developed models met the required accuracy for antibiotic concentration prediction.

Conclusions:

• Machine learning, specifically GRNN-PSO, offers a viable indirect method for predicting antibiotic concentrations in water.
• This approach can aid in water quality management and environmental monitoring.
• Further research can validate this method for diverse aquatic environments and antibiotic types.