Chemical Authentication of Extra Virgin Olive Oil Varieties by Supervised Chemometric Procedures

• Published in: Journal of agricultural and food chemistry Vol. 50; no. 3; pp. 413 - 418
• Main Authors: Bucci, Remo, Magrí, Andrea D, Magrí, Antonio L, Marini, Domenico, Marini, Federico
• Format: Journal Article
• Language: English
• Published: Washington, DC American Chemical Society 30-01-2002
• Subjects: Biological and medical sciences; Discriminant Analysis; Fat industries; Food industries; Fundamental and applied biological sciences. Psychology; Italy; Neural Networks (Computer); Oleaceae - classification; Olive Oil; Plant Oils - analysis; Quality Control; Topography, Medical; Italy; Discriminant analysis; Oleaceae; Olea europaea; Chemical method; Olive oil; Pattern recognition; Neural network; Fatty acids; Gas chromatography; Dicotyledones; Authentication; Angiospermae; Quality control; Analytical method; Spermatophyta; Chemical composition; Intraspecific comparison; Oil plant (vegetal); Cultivar
• DOI: 10.1021/jf010696v
• EISSN: 1520-5118
• ISSN: 0021-8561

This work has focused on discriminating extra virgin olive oils from Sabina (Lazio, Italy) by olive fruit variety (cultivar). A set of oils from five of the most widespread cultivars (Carboncella, Frantoio, Leccino, Moraiolo, and Pendolino) in this geographical area was analyzed for chemical composition using only the Official Analytical Methods, recognized for the quality control and commercial classification of this product. The obtained data set was converted into a computer-compatible format, and principal component analysis (PCA) and a method based on the Fisher F ratio were used to reduce the number of variables without a significant loss of chemical information. Then, to differentiate these samples, two supervised chemometric procedures were applied to process the experimental data:  linear discriminant analysis (LDA) and artificial neural network (ANN) using the back-propagation algorithm. It was found that both of these techniques were able to generalize and correctly predict all of the samples in the test set. However, these results were obtained using 10 variables for LDA and 6 (the major fatty acid percentages, determined by a single gas chromatogram) for ANN, which, in this case, appears to provide a better prediction ability and a simpler chemical analysis. Finally, it is pointed out that, to achieve the correct authentication of all samples, the selected training set must be representative of the whole data set. Keywords: Olive oil; pattern recognition; linear discriminant analysis; artificial neural network