Classification of the fragrance properties of chemical compounds based on support vector machine and linear discriminant analysis

Classification models of the fragrance properties of chemical compounds were performed using linear and non-linear models. The dataset was divided into three classes on the basis of their fragrances: apple, pineapple and rose. The three-class problem was first explored by a linear classifier approach, using linear discriminant analysis (LDA). A more accurate prediction model, the non-linear machine-learning technique, support vector machine (SVM), was subsequently investigated. Descriptors calculated from the molecular structures alone were used to represent the characteristics of compounds. The model containing four descriptors founded by SVM showed better predictive ability than LDA. The accuracy in the prediction for the three datasets was 96.6%, 80.0% and 100% for SVM, respectively. The results indicate that SVM can be used as a powerful modelling tool for QSAR studies and the selected descriptors can represent the fragrances of these chemical compounds. Copyright (C) 2008 John Wiley & Sons, Ltd.