Prediction of density for primary alcohols mixtures using linear and nonlinear methods

Multiple linear regression (MLR) and artificial neural network models (ANN) were successfully developed for predicting of density for primary alcohol mixtures. A large number of descriptors were calculated with Dragon software and a subset of calculated descriptors was selected from 18 classes of Dragon descriptors with a stepwise multiple linear regression as a feature selection technique. 5 calculated descriptors and temperature were selected as the most feasible descriptors. After training and optimization of the ANN architectures and parameters, the performance of the model was investigated by the test set. The results obtained using ANN model were compared with the experimental values as well as with those obtained using multiple linear regression (MLR) model. The correlation coefficient (R2) and mean squared error (MSE) for test set with MLR model were 0.9449, 5.091 and those for ANN model were 0.9827 and 4.688, respectively. These results showed the superiority of ANN model over MLR model.