Mehdi Hatami

The main aim of the present work was development of a quantitative structure-property relationship (QSPR) method using an artificial neural network (ANN) for the prediction of inherent viscosity (h inh) of a data set of 75 optically active polymers containing natural amino acids. The total of 540 descriptors was calculated for all molecules in the data set. In the next step an ANN was constructed and trained for the prediction of h inh of polymers. The inputs of this neural network are theoretically derived descriptors that were chosen by genetic algorithm (GA) and multiple linear regression (MLR) feature selection techniques. The values of standard errors for the neural network calculated h inh of training, test and validation sets are 0.023, 0.030 and 0.031, respectively. Comparison between these values and other statistical values reveal the superiority of the ANN model over the MLR one