Linear support vector machine to classify the vibrational modes for complex chemical systems

Abstract

Classification of vibrational modes into hindered internal rotation (HIR) and harmonic oscillation modes is important to obtain correct thermodynamic data for a chemical species for a wide range of temperatures. In this study, we propose a multivariate linear support vector machine (SVM) model to solve this challenging binary classification problem. The results of the proposed model were found to be similar to those of logistic regression and 2-5% better than those of the rule-based method. Moreover, the number of features found by linear SVM was also fewer than that of logistic regression (five versus six), which makes it easier to be interpreted by chemists. The detailed explanation of such differences is also presented. The three models were implemented in the GUI of the Multi-Species Multi-Channel Software Suite (Duong et al., Int. J. Chem. Kinet, 2015, 564) to facilitate the determination of HIR modes as well as the calculation of thermodynamic properties for a chemical species of interest.