Machine learning-enhanced fluorescence spectroscopy for the quality assessment of extra virgin olive oil during ageing

The safeguarding of extra virgin olive oil (EVOO) quality throughout its shelf life is imperative due to its vulnerability to quality deterioration caused by auto-oxidation and photo-oxidation. This study investigates machine learning (ML) capabilities applied to fluorescence spectroscopy to detect the ageing of EVOO. The quality of EVOO was assessed by UV-absorption spectroscopy measurements as mandated by European Regulations. In parallel, excitation-emission matrices (EEMs) were measured to determine the predictive potential of ML approaches applied to fluorescence data. First, two excitation wavelengths (480 nm and 300 nm) are identified as exhibiting the maximum relative change in fluorescence intensity, serving as potential indicators of EVOO ageing. Then, ML algorithms were developed to predict olive oil quality based on highly aggregated spectral data at these excitation wavelengths. The algorithms successfully identify still good EVOOs from aged EVOOs with over 90% accuracy, proposing an innovative approach that foregoes the need for detailed chemical analysis. This work shows the potential of ML-based approaches applied to fluorescence to replace traditional, labourintensive analyses. Therefore, it paves the way for the development of a compact, field-deployable fluorescence sensing device for rapid and objective quality control in olive oil production and early detection of oxidation or adulteration, and aiding in the classification of olive oils for market purposes.