Artificial Intelligence Models for Efficiency Estimation of Adsorbents in Rare-Earth Element Recovery Based on Feature Engineering

Efficiently separating rare-earth elements (REEs) remains a significant challenge due to selectivity, costs, and potential environmental pollution. This research presents an innovative approach to the efficiency of adsorbents in REE recovery from contaminated water based on feature engineering using machine-learning algorithms and the RDKit toolkit. By analyzing a comprehensive data set from experimental articles, the influential features of promising adsorbents were optimized. The importance of each input feature on the target label ( Q ad ) reveals that the molecular weight of the first functional group significantly impacts adsorption efficiency. The study delves into electron configurations, atomic properties, and thermodynamics, emphasizing the need for balanced energy states, binding affinities, and various bonding mechanisms at play. This accurate model ( R 2 values of 0.928 for both training and testing) provides valuable insights into estimating adsorbent efficiency for REE recovery, paving the way for sustainable materials and promoting novel adsorbent practices in the industry.