Abstract
The potential of Raman spectroscopy for honey authenticity control purposes was investigated with respect to its geographical and botanical origin. For this aim, authentic honey samples from Romania and France were employed in this study. In this regard, two types of processing approaches were used, Soft independent modeling class analogy (SIMCA) and Machine Learning (ML) algorithms. A correlation between SIMCA classification and ML prediction model was observed for honey variety and geographical discrimination. Thus, it appears that when ML algorithms misclassified mono-varietal honeys, SIMCA data treatment provided a low interclass distance revealing a low ability of the model to discriminate among classes. A similar correlation between SIMCA and ML results was obtained for geographical classification, even if SIMCA model apparently provided a better classification of Romanian honeys. However, the obtained interclass distance, lower than unit, revealed that the discriminant information contained in Raman spectra is better linked with varietal composition than to geographical origin.
Published Version
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call