Prediction of viscosity of imidazolium-based ionic liquids using MLR and SVM algorithms

Abstract

In this work, two models, one integrating the fragment contribution-corresponding states (FC-CS) method with multiple linear regression (MLR) algorithm and another. With support vector machine (SVM) algorithm, are proposed to predict the viscosity of imidazolium-based ionic liquids (ILs). The FC-CS method is applied to calculate the pseudo-critical volume and compressibility factor (Vc and Zc) as well as the boiling point temperature (Tb) which are employed to predict the viscosity with the MLR and SVM algorithms. A large data set of 1079 experimental data points of 45 imidazolium-based ILs covering a wide range of pressure and temperature is applied to validate the two models. The average absolute relative deviation (AARD) of the entire data set of the MLR and SVM is 24.2% and 3.95%, respectively. The nonlinear model developed by the SVM algorithm is much better than the linear model built by the MLR, which indicates the SVM algorithm is more reliable in the prediction of the viscosity of imidazolium-based ILs.