Premium
Stepwise geographical traceability of virgin olive oils by chemical profiles using artificial neural network models
Author(s) -
GarcíaGonzález Diego L.,
Luna Guadalupe,
Morales Maria T.,
Aparicio Ramón
Publication year - 2009
Publication title -
european journal of lipid science and technology
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.614
H-Index - 94
eISSN - 1438-9312
pISSN - 1438-7697
DOI - 10.1002/ejlt.200900015
Subject(s) - traceability , olive oil , artificial neural network , identification (biology) , edible oil , test set , mathematics , artificial intelligence , pattern recognition (psychology) , computer science , chemistry , food science , statistics , biology , botany
The geographical traceability of virgin olive oils implies the use of analytical methods that allow the identification of the origin of the oil and the authentication of the information boasted on the labels. In this work, the geographical identification of the virgin olive oils has been addressed by complete chemical characterisation of samples (64 compounds analysed by GC and HPLC) and the design of artificial neural network (ANN) models for each one of the levels of a proposed classification scheme. A high number of samples (687) from Spain, Italy and Portugal served as training and test sets for the ANN models. The highest classification level, focused on the grouping of samples by country, was achieved through analysis of fatty acids, with 99.9% of samples classified. Other levels (region, province, Protected Designations of Origin or PDO) were focused on Spanish oils and required additional series of compounds (sterols, alcohols, hydrocarbons) as well as the fatty acids to obtain classification rates higher than 90%. The classification of oils into different PDOs – the last and most difficult level of classification – showed the highest root mean square errors. The classification percentages, however, were still higher than 90% in the test set, which proves the application of the traceability methodology for a chemical verification of PDO claims.