Prediction of the Lee retention indices of polycyclic aromatic hydrocarbons by artificial neural network

Abstract

A quantitative structure retention relationship technique using an artificial neural network (ANN) has been used for the prediction of the Lee retention indices for PAHs on SE-52 and DB-5 stationary phases. The selected descriptors that appear in the ANN model are the boiling point, molecular weight, connectivity index and the Schabron molecular size descriptor. The network was trained and optimized using a training and validation data sets. For the evaluation of the predictive power of the ANN, the optimized network was used to predict the temperature-prograrnmed Lee retention indices of two unseen testing data sets. The results obtained showed that the mean of relative errors and the correlation coefficients between the calculated ANN and the experimental values of Lee retention indices for the validation and two testing sets are 1.42% and 0.9460 on SE-52; 1.32% and 0.9381; 1.43% and 0.8939 on DB-5 stationary phases, respectively. These values are in good agreement with the relative error obtained by experiment.