Optimal parametrization of structure for prediction of properties of alkanes^†

Abstract

A generalised statistical model has been developed using algorithmically derived independent variables based on chemical structure for quantitative prediction of various physical properties of some hydrocarbons. Out of the fourteen principal components generated from principal component analysis (PCA) of the 14 topological parameters, by the use of only four principal components, the physical properties boiling point, molar volume, molecular refraction, heat of vaporisation, critical temperature and surface tension have been predicted reasonably well.