Machine-learning approach for prediction and analysis of quantitative and qualitative parameters of binary polar liquids

Abstract

Quantitative and qualitative parameters are essential for comprehending the intermolecular interactions in binary polar liquids. In this work, complex permittivity and excess dielectric constant of alcohol–water mixture is used for the quantitative and qualitative analysis, respectively. Aqueous solutions of methanol, ethanol, propanol and isopropyl alcohol are considered. The frequency dispersion of permittivity and the intricate structure of these liquids make the analysis a difficult task. A decision tree regression-based machine-learning model is proposed for the prediction of parameters. For quantitative analysis, the dataset is prepared by measuring the complex permittivity of the mixture using dielectric probe kit–N1501A of Keysight Technologies over a frequency range of 0.2–20 GHz at 25°C. For qualitative analysis, available standard equations are modified to calculate the excess dielectric constants in the specified frequency range. The proposed model requires only three input parameters—frequency, volume fraction of alcohol and static dielectric constant of alcohol to make the prediction. Performance comparison of the model with the measured values of complex permittivity shows minimum error. The analysis reveals the effect of the volume fraction of alcohol and frequency on complex permittivity and excess dielectric constant of the mixture. The proposed model is a novel and reliable prediction tool that can be used for both quantitative and qualitative analysis of alcohol–water mixture.