Geographical classification of olive oils by the application of CART and SVM to their FT-IR

This paper reports the application of Fourier-transform infrared (FT-IR) spectroscopy to the geographical classification of extra virgin olive oils. Two chemometrical techniques, classification and regression trees (CART) and support vector machines (SVM) based on the Gaussian kernel and the recently introduced Euclidean distance-based Pearson VII Universal Kernel (PUK), were applied to discriminate between Italian and non-Italian and between Ligurian and non-Ligurian olive oils. The PUK is applied in literature with success on regression problems. In this paper the mapping power of this universal kernel for classification was investigated. In this study it was observed that SVM performed better than CART. SVM based on the PUK provide models with a high selectivity and sensitivity (thus a better accuracy) as compared to those obtained using the Gaussian kernel. The wave numbers selected in the classification trees were interpreted demonstrating that the trees were chemically justified. This study also shows that FT-IR spectroscopy associated with SVM and CART can be used to correctly discriminate between various origins of olive oils, demonstrating that the combination of techniques might be a powerful tool for supporting the claimed origin of olive oils. Copyright (c) 2007 John Wiley & Sons, Ltd.