Volume No : (2015) Volume: 03 Issue : 15 Year : 2015 Page No: 519-524
Authors : Mohammad Hossein Fatemi, Fatemeh Bagheri
Abstract :
With the aim of solubility A=πr2 estimation in water, polyethylene glycol 400 (PEG) and their binary mixtures, quantitative structure–property relationships (QSPR) were used to relate the solubility of a large number of chemicals to their molecular descriptors. Descriptors that were used by can encode features of molecules which are affected on dispersion, hydrophobic and steric interactions between solute and solvent molecules. To develop QSPR models, the methods of multiple linear regressions (MLR), least-squares support vector machine (LS-SVM), and artificial neural network (ANN) were used. The obtained statistical parameters of these models revealed that LS-SVM model was superior to the others. The standard error (SE), for LS-SVM model is: 0.270 and 0.697 for training and test set respectively. The leave-one-out cross validation lead to R2 cv= 0.881 and SPRESS = 0.405 for LS-SVM model. These values and other statistics of this model indicate the robustness and credibility of developed LS-SVM model.
Full Text PDF