Predictive modeling of physicochemical properties and ionicity of ionic liquids for virtual screening of novel electrolytes

In this paper, we present three regression models developed for predicting electrical conductivity, viscosity, and density of ionic liquids. Combining these machine learning models and our previously developed models for prediction of the IL thermal properties enables virtual screening of new ILs with desired properties. Three different cross-validation protocols are applied to test the performance of the models and their predictive power is discussed. A simple classification model is developed to estimate the ionicity of ILs in terms of the Walden plot. The Walden products of new theoretical ILs are analyzed in terms of their deviation from the "ideal KCl line" and structural fragments, which influence the ionicity most significantly, are distinguished. The experimental data and all the developed models for predicting the electrical conductivity, viscosity, density, and ionicity of ILs, can be found at https://ochem.eu/article/158738. To the best of our knowledge, the models presented here are the first open access ones.