Industrial wastewater identification based on HPLC combined with data standardization and ensemble learning...

Area of Science:

• Environmental Science
• Analytical Chemistry
• Data Science

Background:

• Industrial wastewater discharges pose significant environmental risks.
• Accurate source identification is crucial for regulatory oversight and environmental protection.

Purpose of the Study:

• To develop a novel approach for discriminating industrial wastewater sources.
• To integrate data standardization with ensemble learning for improved accuracy.

Main Methods:

• High-performance liquid chromatography (HPLC) for sample collection.
• Local stretching alignment and Gaussian fitting for chromatographic data peak alignment.
• Comparison of Random Forest (RF) and Extreme Gradient Boosting (XGBoost) ensemble learning algorithms.
• Hyperparameter optimization using the Optuna framework.
• Model evaluation via five-fold cross-validation.

Main Results:

• The optimized RF model achieved 97.87% average cross-validation accuracy and 98.28% test set accuracy.
• An F1 score of 0.9799 and 95.08% accuracy on newly collected samples were obtained.
• The approach demonstrated excellent performance and a balanced precision-recall trade-off.

Conclusions:

• The proposed integrated approach offers an efficient and reliable tool for tracing industrial wastewater sources.
• This method enhances capabilities for environmental monitoring and regulatory enforcement.
• Accurate source identification is key to mitigating the environmental impact of industrial activities.