A new method of Ionic Fragment Contribution-Gradient Boosting Regressor for predicting the infinite dilution activity coefficient of dichloromethane in ionic liquids

Abstract

Ionic liquids (ILs) have shown huge potential advantages as solvents to absorb and recover dichloromethane (DCM) from waste gases. The infinite dilution activity coefficient ( γ ∞ ) of DCM in ILs is an important parameter, which can be used to predict the vapor-liquid equilibrium of DCM-IL systems. In this work, a new model of calculating the γ ∞ of DCM in ILs is established based on ionic fragments contribution (IFC) and gradient boosting regressor (GBR) algorithm. IFC is used to obtain the surface charge density distribution area of ILs (S σ-profile ) that is the input of GBR. GBR is used to learn the mapping relationship between input feature and γ ∞ of DCM in ILs. The database of the γ ∞ of DCM in ILs composed of 29 cations and 22 anions includes 72 experimental data and 421 COSMO calculation data, which was employed to establish the IFC-GBR model and predict the γ ∞ of DCM in ILs. The coefficient of determination (R 2 ) and mean absolute error (MAE) of the IFC-GBR model test set are 0.9703 and 0.0519, respectively. Also, this model has excellent generalization capability of predicting evidenced by high 10-fold cross-validation coefficients of determination in the range 0.9474–0.9481. These results indicate that the proposed model can accurately predict γ ∞ of DCM in ILs, then provide the important data for developing a new process of absorbing and desorbing DCM by IL-based technologies.