A Comparison of “Neural Networks and Multiple Linear Regressions” Models to Describe the Rejection of Micropollutants by Membranes

Abstract

A rejection process of organic compounds by nanofiltration and reverse osmosis membranes was modelled using the artificial neural networks. Three feed-forward neural networks based on quantitative structure-activity relationship (QSAR-NN models) characterised by a similar structure (twelve neurons for QSAR-NN<sub>1</sub>, QSAR-NN<sub>2</sub>, and QSAR-NN<sub>3</sub> in the input layer, one hidden layer and one neuron in the output layer), were constructed with the aim of predicting the rejection of organic compounds. A set of 1394 data points for QSAR-NN<sub>1</sub>, 980 data points for QSAR-NN<sub>2</sub>, and 436 data points for QSAR-NN<sub>3</sub> were used to construct the neural networks. Good agreements between the predicted and experimental rejections were obtained by QSAR-NN models (the correlation coefficient for the total dataset were 0.9191 for QSAR-NN<sub>1</sub>, 0.9338 for QSAR-NN<sub>2</sub>, and 0.9709 for QSAR-NN<sub>3</sub>). Comparison between the feed-forward neural networks and multiple linear regressions based on quantitative structure-activity relationship “QSAR-MLR” revealed the superiority of the QSAR-NN models (the root mean squared errors for the total dataset for the QSAR-NN models were 10.6517 % for QSAR-NN<sub>1</sub>, 9.1991 % for QSAR-NN<sub>2</sub>, and 5.8869 % for QSAR-NN<sub>3</sub>, and for QSAR-MLR models they were 20.1865 % for QSAR-MLR<sub>1</sub>, 19.3815 % for QSAR-MLR<sub>2</sub>, and 16.2062 % for QSAR-MLR<sub>3</sub>).