Artificial Neural Network Models for the Prediction of Refractive Index of Pure Alcohols

In this study, the refractive index of 9 pure samples of 1-alcohols were experimentally investigated at (22 °C and 25 °C). These data were used to establish two different neural network models of multi-layer perceptron (MLP) and radial basis function (RBF)for prediction of refractive indexes of pure alcohols. For this purpose the temperature, molecular mass, and functional groups of the compounds were considered as the input parameters and the refractive index was considered as the only output of the two neural networks. 70% of the data are given as the training data, 15% as the validation data and 15% as the test data. The optimized MLP neural network had the mean square error (MSE) of 0.00000483712 for the training data, 0.0000649641 for the validation data, 0.000011277 for test data and the correlation coefficient (R) of 0.99238 with 6 optimal neurons in the hidden layer. Also in the optimized RBF neural network with 15 neurons had the minimum MSE of 0.0000135835. Keywords - Refractive index, Artificial Neural Network, alcohol, optimization, volumetric percentages