Machine Learning for Accurate Prediction of Henry’s Law Constant in CO₂–Ionic Liquid Systems

To accurately and reasonably describe the influence of molecular structure, charge characteristics, and other intrinsic properties on Henry’s Law Constant (HLC), we have designed the Multiscale Group Contribution Descriptor-Machine Learning (MGC-ML) model based on the revised group contribution method─the weighted group contribution. This method extracts first-order descriptors using the group contribution approach and then utilizes a machine learning model to describe the impact of molecular structure and electronic characteristics on the target properties. Experimental results demonstrate that this method not only ensures theoretical reliability but also exhibits exceptionally high predictive accuracy with a smaller amount of data and fewer molecular features. Considering that this model combines chemical principles with data-driven approaches, it is likely to be proven reasonable and effective in addressing more general objectives in complex systems.