A combined classification modeling strategy for detection and identification of extra virgin olive oil adulteration using Raman spectroscopy

Abstract

Identifying and discriminating of pure edible oils as well as detecting and identifying the type of adulteration, are essential issues in the field of food-related research. Extra virgin olive, hazelnut, canola, soya, and sunflower oils were investigated by Raman spectroscopy to answer two main questions: How can it be recognized that an unknown sample of edible oil belongs to one of these five types of pure oil? Also, if the unknown sample is adulterated extra virgin olive oil, how can the type and composition of that sample be recognized? The problem-based research was planned by proposing a combined classification strategy for achieving the proper solutions. In this approach, class modeling was combined with discriminant analysis to benefit from the advantages of both methods. In the first step, the Raman spectra of five pure edible oil classes were used to model the target objects by DD-SIMCA to exclude adulterated samples. Then PLS-DA was used for the discrimination of five edible oils classes. This strategy was successfully applied for several binary classes, including different types of extra virgin olive oil adulterations with different compositions. Raman spectroscopy, accompanied by the proposed combined classification strategy, showed good power in solving the pattern recognition problems in food adulteration subjects. The proposed combined classification strategy allowed the discrimination between pure edible oil samples with up to 100% of correct classifications. The developed method showed also high performances based on the figures of merit (up to 100% sensitivity and specificity) for the detection and identification the type of adulteration in extra virgin olive oil.