A combination of attenuated total reflectance–Fourier transform infrared (ATR–FTIR) spectroscopy and chemometric techniques was used to classify different trademarks of edible oils available on the Turkish markets. A total of 144 spectra of edible oil samples, including extra virgin olive oil (VOO), hazelnut oil (HNO), cottonseed oil (CSO), sunflower oil (SFO) and soybean oil (SBO), was recorded. The feasibility of ATR–FTIR with multivariate data analysis for discrimination of extra VOOs from other edible oils was also evaluated. Classification of edible oils was performed using principal components analysis (PCA), hierarchical cluster analysis (HCA), linear discriminant analysis (LDA) and soft independent modeling of class analogies (SIMCA). The spectra collected from wavelength region of 4000–650 cm-1 and 28 different wavelength ranges selected from full spectra were evaluated for optimal classification models. All multivariate analysis provided excellent discriminations between the edible oil classes with low classification error. LDA models constructed with five predictors, and a total of 100% of edible oil samples from different trademarks were correctly classified. Furthermore, no misclassification was reported for the discriminant analysis in supervised SIMCA models with an accuracy of 95%. Consequently, ATR–FTIR spectroscopy combined with multivariate data analyses provides excellent illustrations of the relative positions of the different brands of commercial edible oils according to their quality and purity.
Classification, Spectroscopy, Chemometrics, Virgin olive oil