Fast prediction of electron-impact ionization cross sections of large molecules via machine learning

The theoretical determination of electron-impact ionization cross section (Q(ion)) for a molecule requires ab initio computation, which is time-consuming for large molecules. We propose a machine learning based method to construct a model for predicting Q(ion) of large molecules without the high-cost ab initio calculation. The model is learned from the data composed of the calculated Q(ion) of the small molecules with fewer constituent atoms and the electron numbers of the corresponding molecules in a train set by a support vector machine. The radial basis function is set as a kernel function to map data to a higher dimensional space. The grid search with 5-fold cross-validation is performed to find optimal hyperparameters in the learning model. The prediction on the test sets composed of CF4, C3F8, SF6, C-6, C6F12, and C6F12O shows that this data-driven model can generate well-agreed Q(ion) and has good generalization performance.