Classification of organic solvents and modelling of their physico-chemicalproperties by chemometric methods using different sets of molecular descriptors

Different sets of molecular descriptors using the k-nearest neighbor classi fication method were used to make a general classification of 152 organic s olvents, the classification being further improved by performing the counte r-propagation artificial neural network. An extensive investigation was mad e of the physico-chemical properties of 152 solvents in a search for quanti tative structure-property relationships (QSPR). Wide sets of molecular desc riptors were tested and regression models were obtained by selecting the be st descriptor subset by genetic algorithm in order to optimize their predic tion power. (C)1999 Elsevier Science B.V. All rights reserved.