Prediction of CO2 solubility in ionic liquids using machine learning methods

Abstract

A comprehensive database containing 10,116 CO2 solubility data measured in various ionic liquids (ILs) at different temperatures and pressures is established. Based on this database, the relationship between CO2 solubility and IL structure, temperature and pressure is correlated using group contribution (GC) methods. Two different machine learning algorithms, namely artificial neural network (ANN) and support vector machine (SVM), are employed to develop the GC models. For the 2023 test-set data, the estimated MAE and R2 are 0.0202 and 0.9836, respectively for the ANN-GC model and for the SVM-GC model they are 0.0240 and 0.9783, respectively. The distributions of prediction errors are plotted for both models to provide more comprehensive knowledge on the model performance. The results indicate that both of the models can give reliable predictions on the CO2 solubilities in ILs and the ANN-GC model performs slightly better than the SVM-based model.