Heavy Metals Potentiometric Sensitivity Prediction by Firefly-Support Vector Machine Modeling Method

The quantitative structure-property relationship (QSPR) method is an efficient and elegant method for estimating the critical parameters of a wide range of compounds. In this work, the QSPR data set included the structures of 45 modified diphenyl phosphoryl acetamide ionophores along with their sensitivity to Cd2+, 2+ , Cu2+, 2+ , and Pb2+. 2+ . The data set was divided into the training set, including 36 compounds, and the test set, including 9 compounds. The stepwise-multiple linear regressions (SW-MLR), firefly multiple linear regressions (FA-MLR), and firefly-support vector machine (FA-SVM) models were produced on the training set with sensitivity of ionophores for Cd2+, 2+ , Cu2+, 2+ , and Pb2+ 2+ for predicting the potentiometric sensitivity of plastic polymer membrane sensors. The FA-SVM model showed good statistical results for all three cations. Internal and external validation was done to ensure the performance of the model. The results showed acceptable accuracy of the proposed method in identifying important descriptors in QSPR. The results of this study and the interpretation of the descriptors entered in the model can help to design new selective ligands.