Developing a Support Vector Machine Based QSPR Model for Prediction of Atmospheric Lifetime of Some Halocarbons

Abstract

Abstract In this work, the atmospheric lifetime of 60 halocarbons was estimated from their theoretical derived molecular descriptors by applying quantitative structure–property relationship (QSPR) methodology. The most relevant descriptors selected by stepwise multiple linear regression were used for developing linear and nonlinear models by using multiple linear regression and support vector machine, respectively. Here we show that the support vector machine model is finely capable for predicting the lifetime of halocarbons. The built support vector machine model was assessed by leave one out cross-validation (Q2 = 0.928, SPRESS = 0.479) and Y-randomization test (R2 = 0.222 for 25 trail) as well as external validation test. The developed support vector machine model was used for prediction of the atmospheric lifetime of some halocarbons with lifetimes not reported experimentally.