Optimizing Cu^2 + adsorption prediction in Undaria pinnatifida using machine learning and isotherm models

Area of Science:

• Environmental Science
• Biotechnology
• Adsorption Science

Background:

• Algae are effective bioadsorbents for heavy metal remediation.
• Traditional adsorption models have limitations in predicting algal bioadsorption capacity.
• Undaria pinnatifida's potential for copper (Cu²⁺) removal is underutilized.

Purpose of the Study:

• To integrate machine learning (ML) with traditional models for predicting Cu²⁺ adsorption by Undaria pinnatifida.
• To determine the relationship between algal bioactive compounds and Cu²⁺ adsorption.
• To identify key factors influencing Cu²⁺ adsorption for optimized environmental remediation strategies.

Main Methods:

• Experimental determination of Cu²⁺ adsorption by Undaria pinnatifida parts.
• Modeling adsorption using Freundlich, pseudo-second-order, and thermodynamic models.
• Application and comparison of ML regression algorithms (CatBoost) for prediction.
• Feature importance analysis using Shapley and Partial Dependence Plots.

Main Results:

• Phlorotannins content positively correlated with Cu²⁺ adsorption capacity.
• Adsorption behavior best described by Freundlich and pseudo-second-order models, indicating multilayer adsorption.
• CatBoost ML model achieved high prediction accuracy (R²=0.9883).
• Initial Cu²⁺ concentration identified as the most significant factor influencing adsorption.

Conclusions:

• The integration of ML and traditional models offers a powerful approach for predicting algal heavy metal adsorption.
• Undaria pinnatifida, particularly its phlorotannin content, shows significant potential for Cu²⁺ remediation.
• ML facilitates accurate predictions, enhancing the targeted utilization of algae in environmental pollution control.