Prediction of selectivity coefficients of univalent anions for anion-selective electrode using support vector machine

Abstract

As a new method, support vector machine (SVM) was applied for the prediction of selectivity coefficients of anion-selective electrode for some univalent anions. In this way a quantitative structure–property relationship model was constructed based on calculated molecular descriptors. In the first step more than 1000 molecular descriptors were calculated. Then the stepwise multiple linear regression method was used to select most important descriptors which are responsible for the selectivity coefficient of anion-selective electrode for interfering anions. The selected descriptors were used as inputs for SVM model. The root-mean-square errors of SVM calculated selectivity coefficients for training and test set are 0.878, 0.890 and the correlation coefficients are 0.95, 0.94, respectively. Also the obtained statistical parameters of cross-validation test on SVM model were Q 2 = 0.858 and SPRESS = 1.050 which revealed the reliability of this model. The results of this study reveal that SVM technique can be used as a powerful tool for prediction of selectivity coefficients of anion-selective electrode from the theoretical derived molecular descriptors.