Advantages of support vector machine in QSPR studies for predicting auto-ignition temperatures of organic compounds

A new method support vector machine (SVM) was first used to develop quantitative structure–property relationship (QSPR) models for predicting auto-ignition temperatures (AIT) of organic compounds. The calibration and predictive ability of SVM were investigated and compared with those of other two common methods, multiple linear regression (MLR) and back-propagation neural network (BPNN). Two different datasets were evaluated. The first one involved a set of 50 alkane compounds, whose structural characteristics were encoded by employing the widely used atom-type E-state indices as molecular descriptors, while the second one dealt with a total of 142 organic compounds, whose molecular structures were described using both the physicochemical parameters and structure descriptors. Both internal and external validations were performed to validate the performance of the resulting models. The results showed that, for both datasets, the calculated AIT values by SVM were in good agreement with the experimental ones, and the performances of the SVM models were comparable or superior to those of MLR and BPNN ones, especially in external predictive ability. This paper provides a new and effective method for predicting AIT of organic compounds, and also reveals that SVM can be used as a powerful chemometrics tool for QSPR studies.