Different types of physical solvents have been utilized for CO2 removal from natural gas in the sweetening process. In this work, quantitative structure-property relationship (QSPR) method is suggested to build powerful models to predict Henry’s law constant (HLC) for CO2 in physical solvents. Modeling the HLC for CO2 as a function of molecular descriptors was achieved by multiple linear regression and descriptor selection was by genetic algorithm. The main proposed model has two simple descriptors, including the number of hydroxyl groups and molecular weight of solvents at fixed temperature. Also, the effect of temperature was studied, and this operational variable was added to the mentioned simple descriptors. In this case, the data set is comprised of 77 HLC for CO2 in solvents and at different temperatures. Several internal and external validation methods demonstrated the excellent ability for prediction, and the average relative deviation of main model was 6.48.