The prediction of the substantivity of anionic, water soluble dyes on a biomass using neural networks

The substantivity of series of anionic, water soluble dyes for a biomass ha s been measured. Using feed-forward back propagation neural networks and re latively simple empirical and semi-empirical molecular descriptors as input variables, the correlation between the predicted substantivity and the obs erved substantivity was determined. Depending on the molecular descriptors used as input variables and the number of hidden layers in the neural netwo rk the correlation between the predicted substantivity and the observed sub stantivity was 0.80 and 0.76. (C) 2000 Elsevier Science Ltd. All rights res erved.