Machine learning prediction of dye adsorption by hydrochar: Parameter optimization and experimental validation.

Area of Science:

• Environmental Science
• Materials Science
• Data Science

Background:

• Global wastewater contamination from dyes necessitates effective removal strategies.
• Hydrochar shows promise as a bio-based adsorbent for dye removal.
• A comprehensive understanding of factors influencing hydrochar dye adsorption is lacking.

Purpose of the Study:

• To explore the complex relationships between hydrochar properties, preparation and experimental conditions, dye types, and adsorption capacity (Q).
• To develop and validate a predictive model for hydrochar dye adsorption using machine learning (ML).
• To create a user-friendly tool for researchers to apply and improve predictive models for sustainable contaminant removal.

Main Methods:

• Utilized twelve distinct machine learning models to assess dye adsorption.
• Employed Gradient Boosting Regressor (GBR) for its superior predictive performance.
• Conducted feature importance analysis and experimental validation to confirm model efficacy.

Main Results:

• The GBR model achieved high accuracy (R² = 0.9629) in predicting dye adsorption.
• Identified experimental conditions and hydrochar properties as the most significant factors influencing adsorption.
• Experimental validation confirmed the GBR model's practical utility (R² = 0.8704).

Conclusions:

• Machine learning effectively enhances the understanding of dye adsorption mechanisms by hydrochar.
• The study provides a robust predictive model and a practical GUI for researchers.
• This work establishes a precedent for using bio-based adsorbents and ML in sustainable wastewater treatment.