Neural network modeling as an efficient approach to predict the density of ionic liquids/ethanol binary systems

Abstract

Ionic liquids (ILs) especially their mixtures are of high interest within the different scientific societies due to their amazing properties. In this regard, a number of attempts have been made to measure, correlate, estimate and calculate the properties of ILs in the neat or mixture forms. Among the different possible predictive methods, artificial neural networks (ANNs) are widely used because of their unique and amazing capabilities for prediction of different parameters. With respect to this paper, a feed-forward ANN model is proposed to model the densities of different binary mixtures of ILs/ethanol. The proposed network is trained and tested with 1078 binary data points gathered by mining into the different published literatures. The data gathered from previously published literatures are separated into two different subsets namely training and testing. The statistical error analysis has shown that the proposed neural network correlated the binary densities with the overall mean absolute percentage error (MAPE), average relative deviation percentage error (ARD%), minimum relative deviation percent (RDmin%), maximum relative deviation Percent (RDmax%) and correlation coefficient ([Formula: see text] of 1.5%, [Formula: see text]0.1%, [Formula: see text]13.0%, 15.0% and 0.9712, respectively.