A Study on Machine Learning Methods' Application for Dye Adsorption Prediction onto Agricultural Waste Activated...

Area of Science:

• Environmental Science
• Materials Science
• Computational Chemistry

Background:

• Dye contamination in wastewater poses significant environmental challenges.
• Agro-wastes offer a sustainable and cost-effective solution for dye adsorption.
• Predictive modeling can optimize the selection and application of adsorbents.

Purpose of the Study:

• To develop and compare machine learning models for predicting dye adsorption capacity of agro-wastes.
• To identify key agro-waste characteristics influencing adsorption efficiency.
• To provide a predictive tool for optimizing wastewater treatment processes.

Main Methods:

• Utilized 350 adsorption experimental datasets for 39 adsorbents, including 16 types of agro-wastes.
• Applied machine learning algorithms: Random Forest (RF), Decision Tree (DT), and Gradient Boosting (GB).
• Performed feature selection to identify the most influential variables for adsorption capacity.

Main Results:

• Random Forest (RF) model achieved the highest accuracy (R² = 0.9) in predicting dye adsorption.
• Five key variables were selected from nine, with pore volume and surface area being most critical.
• Agro-waste characteristics contributed 50.7% to adsorption efficiency, with particle size having minimal impact.

Conclusions:

• Machine learning models, particularly RF, can accurately predict agro-waste adsorption capacity.
• Understanding key agro-waste characteristics guides the development of effective dye removal strategies.
• Predictive modeling reduces experimental efforts and directs research towards optimal adsorbent materials for wastewater treatment.