Abstract Extra Virgin Olive Oil (EVOO) is globally valued for its exceptional sensory qualities and health benefits. However, its high market value and demand make it a target for adulteration, particularly with lower-cost seed oils, posing significant socioeconomic and health concerns. Current official methods for detecting adulteration rely heavily on chromatographic techniques, which are labor-intensive, environmentally unfriendly, and unsuitable for real-time applications. This study proposes a novel methodology integrating Near-Infrared (NIR) spectroscopy with advanced machine learning algorithms to address these limitations. The methodology was tested on EVOO samples adulterated with five common seed oils (corn, sunflower, high-oleic sunflower, rapeseed, and soybean) at varying concentrations. The NIR spectra of the samples were processed using Principal Component Analysis and analyzed with multiple classification models, including Decision Tree, Linear Discriminant Analysis, Naive Bayes, Support Vector Machine, and K-Nearest Neighbors, which consistently achieved the best performance in terms of accuracy, precision, recall, and F1-score. The proposed system demonstrated high identification rates (up to 96%) and efficient quantification of adulterants with a mean prediction error of 1.2%. This approach offers a rapid, cost-effective, and environmentally sustainable alternative to traditional methods, with potential for real-time, online implementation in industrial settings. Future research will focus on enhancing data preprocessing workflows and developing standardized calibration models to further optimize this methodology for broader adoption in the olive oil industry.
Gila et al. (Sat,) studied this question.