A review on created QSPR models for predicting ionic liquids properties and their reliability from chemometric point of view

Ionic liquids (ILs) are chemicals that have attracted the attention of scientists and engineers because of their various applications in industry. It is important to predict the value of their properties. A useful tool for predicting ILs properties is quantitative structure-property relationship (QSPR). Various QSPR models have been created by different research groups for estimating the properties of ILs. In this research, created QSPR models by researchers have been reviewed to obtain a good view about their weak and strong points. It seems the lack of large datasets for some properties has led to unreliable QSPR models. There are QSPR models with high predictive power and wide applicability domain for properties such as melting point, viscosity, density and surface tension. Unfortunately, in developing some QSPR models, several mistakes have occurred. For instance, some of them have not been validated against a test set, many variables have been used in several QSPR models and chance correlation, collinearity and multi-collinearity have not been investigated for some of them. These factors make them unreliable QSPR models for predicting the properties of ionic liquids. Also, in many cases, datasets have small size and the applicability domain of QSPR models have not been determined. (C) 2018 Elsevier B.V. All rights reserved.