Optimizing Cu adsorption prediction in Undaria pinnatifida using machine learning and isotherm models.
Journal:
Journal of hazardous materials
Published Date:
Apr 7, 2025
Abstract
Algae are cost-effective bioadsorbents for heavy metal remediation, yet their potential is underutilized due to limitations in traditional adsorption models. This study integrates machine learning (ML) techniques with traditional models to predict the Cu adsorption capacity by Undaria pinnatifida, enabling more efficient and targeted strategies for heavy metal removal. The study determined the relationship between bioactive compounds (mannitol, alginate, phlorotannins) content in different parts (blade, stipe, sporophyll) of algae and revealed a positive correlation between phlorotannins and Cu²⁺ adsorption capacity. The adsorption behavior of algal blades was best described by the Freundlich model (R=0.9858), pseudo-second-order kinetic model (R=0.9989), and thermodynamic model (R=0.9912). These models suggest multilayer adsorption and confirm the spontaneous nature of the adsorption process. ML regression using factors such as temperature, initial concentration, time, and equilibrium concentration, with CatBoost providing the best predictions (R=0.9883). Feature importance analysis (Shapley and Partial Dependence Plot) identified the initial concentration as the most influential factor affecting Cu adsorption. This study presents a novel approach by combining traditional models and ML techniques to predict algal Cu adsorption capacity. The findings highlight the potential of ML for accurate predictions and provide valuable insights for enhancing the utilization of algae in environmental pollution control.
