Prediction of Equilibrium Constants in Aqueous Solution I. The Extrapolation of Equilibrium Constants to Zero Ionic Strength Using PLS, Artificial Neural Networks, and Genetic 'Soft' Modelling

摘要

Extrapolation of formation constants to zero ionic strength using "soft" modelling with partial least-squares, genetic algorithm, and artificial neural networks (ANN) methods was examined and results of individual approaches were compared. The methods allow a rapid and sufficiently accurate prediction of thermodynamic formation constants, ion-size parameters, and salting-out coefficients from experimental equilibrium data, among them the ANN method was found most reliable.