Agro-Climatic Information to Enhance the Machine-Learning Classification of Olive Oils from Near-Infrared Spectra

Abstract

The integrity of extra virgin olive oil (EVOO) quality markers can be compromised owing to deceptive marketing practices, such as misleading geographical origin claims or counterfeit certification labels, i.e., protected designations of origin (PDO). Therefore, it is imperative to introduce ecofriendly, rapid, and economical analytical methods for authenticating EVOO, such as near-infrared (NIR) spectroscopy. Unlike traditional techniques such as chromatography, NIR spectra contain unresolved bands; hence, chemometric tools such as principal component analysis (PCA) are essential for extracting valuable information from them. Herein, PCA was employed to reduce the high dimensionality of the NIR spectra. The PCA factors were then integrated as explanatory variables in machine-learning classification models, enabling the classification of EVOO based on its geographical origin or PDO. Furthermore, the classification models were improved by incorporating agro-climatic data, resulting in a noticeable improvement in the accuracy and reliability of the results. These results were cross-validated by changing the calibration and validation subsamples in successive iterations and averaging the obtained ratios. The results were robust when the olive varieties differed. Consequently, our findings highlight the potential benefits of incorporating agro-climatic information with NIR spectral data in classification models.