FTIR spectroscopy combined with machine learning for the classification of Mediterranean honey based on origin

This study focused on the differentiation of Mediterranean honeys based on their geographical and botanical origin using FTIR-ATR spectroscopy combined with chemometrics. A total of 156 commercial honey samples, classified as thyme, pine, or polyfloral, were gathered from five Mediterranean countries, namely Greece, Malta, Spain, Tunisia, and Turkey. Melissopalynological and physicochemical analyses were performed to characterize the honey samples. The geographical and botanical origins were identified using Principal Component Analysis (PCA) in conjunction with Random Forest (RF) and Data-Driven Soft Independent Modeling of Class Analogies (DD-SIMCA). The analysis utilized the spectral range of 1800 – 750 cm -1 , preprocessed with the first derivative. Both one-class (DD-SIMCA) and multiclass (RF) classification techniques demonstrated high accuracy, exceeding 90% in most cases. Specifically, the best results in terms of differentiation of geographical origin using all samples were achieved by DD-SIMCA, yielding over 95% accuracy for all countries, with the exception of Tunisia with an accuracy of 87%. These findings highlight the robust predictive potential of FTIR-ATR spectroscopy combined with chemometric methods for determining both the geographical and botanical origins of honey. This methodology provides a fast, non-destructive tool for verifying the origin of Mediterranean honey, contributing to improved food traceability and supporting the honey industry. • Honey samples analyzed from Greece, Malta, Spain, Tunisia, and Turkey. • RF and DD-SIMCA used for origin identification with high accuracy. • Classification accuracy exceeded 90% for most origin determinations. • FTIR spectroscopy can determine the geographical and botanical origin of honeys. • The proposed method ensures fast, non-destructive honey origin verification.

Перевод пока недоступен