Abstract
We developed two distinct regression models based on machine learning approach for estimating CO2 loading in solvents in this study. The methods of Adaptive boosted support vector (SVR) machines and the Adaptive boosted Gaussian process (GPR) are the models employed in this study for approximating carbon dioxide loading in the solvent for the purpose of environmental applications. Indeed, the case study is molecular separation for capture of CO2 from a mixture using amino acid salt solutions and understanding the influence of variables on variation of gas solubility. The target output parameter in the computations is the CO2 solubility in liquid phase (alpha). The scores of the two models of boosted SVR and boosted GPR were 0.830 and 0.991, respectively, using the R-2 criterion. Furthermore, the error rate with the MAPE standard is 2.21758E-01 and 1.06938E-01 for the two models, respectively. The boosted SVR and regression of the boosted GPR with the MAE criterion have error rates of 1.64937E01 and 7.28501E-02, respectively, as the models' third efficiency metric. The used machine learning models indicated to be robust for simulation of CO2 absorption in liquid phase for environmental applications and can be used to save time and costs of measurements. (C) 2022 Elsevier B.V. All rights reserved.