A Machine Learning-Based Modeling Approach for Dye Removal Using Modified Natural Adsorbents.

Area of Science:

• Environmental Science
• Materials Science
• Computational Chemistry

Background:

• Wastewater contamination by cationic dyes poses environmental risks.
• Developing sustainable and cost-effective dye removal methods is crucial.
• Agricultural waste offers a promising source for novel biosorbent materials.

Purpose of the Study:

• To evaluate the efficacy of fruit seed-derived biosorbents for methylene blue (MB) dye removal.
• To investigate the influence of various experimental parameters on dye adsorption.
• To apply machine learning models for predicting biosorbent performance.

Main Methods:

• Levulinic acid (LA)-modified almond, apricot, and walnut shells were synthesized as biosorbents.
• Experimental data (105 points) were collected varying pH, adsorbent dose, concentration, time, and temperature.
• Machine learning algorithms including Gradient Boosting (GB), MLP, XGBoost, and Random Forest were employed for regression analysis.
• Spearman correlation was used to rank attribute importance due to non-normal data distribution.

Main Results:

• The Gradient Boosting (GB) model demonstrated superior performance in predicting MB removal.
• The GB model achieved R² values of 0.8858 for removal percentage and 0.9532 for adsorption capacity.
• Key parameters influencing adsorption were identified through correlation analysis.

Conclusions:

• Fruit seed waste, modified with LA, serves as an effective biosorbent for MB dye removal.
• Machine learning models, particularly GB, accurately predict biosorbent performance.
• This approach offers a sustainable solution for dye wastewater treatment.