基于支持向量机方法的烷烃辛烷值预测

Prediction of octane numbers of alkanes based on support vector machine

基于定量结构-性质相关(QSPR)原理,研究化学物质的结构与性能之间的关系,应用支持向量机(SVM)回归方法,建立了根据分子结构预测烷烃马达法辛烷值的数学模型,分别采用内部和外部验证的方式对模型性能进行了验证。结果表明,模型具有较高的稳定性以及预测能力。该方法的提出为工程上提供了一种根据分子结构有效预测烷烃马达法辛烷值的新方法。

The relationships between chemical substance structures and properties were investigated based on the quantitative structure-property relationship（QSPR） studies.The support vector machine（SVM）was employed to build the model for prediction of the motor octane number（MON） of pure alkane based on its molecular structure.Both internal and external validations were performed to validate the performances of the resulting model.The results showed that the model had good robustness and high prediction ability which provides a new method for predicting MON of alkanes in engineering through the molecular structure only.