A “non-linear” quantitative structure-property relationship for the prediction of electrical conductivity of ionic liquids

In this communication, a “non-linear” quantitative structure-property relationship is proposed to represent/predict the electrical conductivity of ionic liquids (ILs). A database comprising 977 experimental data for 54 ILs was collected from the literature. Computation of the molecular descriptors for each IL from their chemical structure of anions and cations was then undertaken. The sequential search algorithm was implemented to select the optimal subset of molecular descriptors containing temperature, 4 anion-based molecular descriptors, and 5 cation-based molecular descriptors. Finally, a least square support vector machine (LSSVM) model was generated using the selected parameters to represent/predict the electrical conductivity of ILs. The proposed model produces a low average absolute relative deviation (AARD) of less than 1.9% taking into consideration all 977 experimental data values.