Application of artificial neural networks for prediction of the retention indices of alkylbenzenes

Artificial neural networks (ANN) with extended delta-bar-delta (EDBD) learn ing algorithms were used to predict the retention indices of alkylbenzenes. The data used in this paper include 96 retention indices of 32 alkylbenzen es on three different stationary phases. Four parameters: temperature, boil ing point, molar volume and the kind of stationary phase, were used as inpu t parameters. These three stationary phases are: PEG, SE-30, SQ. The 96 gro up data were randomly divided into two sets: a training set (including 64 g roup data) and a testing set (including 32 group data). The structures of n etworks and the learning times were optimized. The best network structure i s 4-7-1. The optimum number of learning time is about 20 000. It is shown t hat the maximum relative error is no more than 3%. The result illustrated t hat the prediction performance of ANN in the field of investigating the ret ention behaviors of alkylbenzenes is very satisfactory. (C) 1999 Elsevier S cience B.V. All rights reserved.