Invited Presentation
RAM
Maria Anastasiadi
Senior Lecturer in Bioinformatics
Cranfield University
Bedford, England, United Kingdom
Mennatullah Shehata
Research Assistant
Cranfield University
Cairo, Al Qahirah, Egypt
Sophie Dodd
Doctoral Researcher
Cranfield University
Cranfield, England, United Kingdom
Zahra Karimi
Doctoral Researcher
Cranfield University
Cranfield, England, United Kingdom
Sara Mosca, PhD
Research Scientist
Central Laser Facility, STFC, Rutherford Appleton Laboratory
Oxford, England, United Kingdom
Pavel Matousek
Professor
STFC
Oxford, England, United Kingdom
Bhavna Parmar
Senior Scientist
FSA
London, England, United Kingdom
Honey is a natural food product, with extreme variability in its physicochemical characteristics mainly depending on its floral and geographical origin. In addition, the global honey supply chain is very complex, and traceability can be challenging to implement, leaving honey susceptible to food fraud malpractices such as origin mislabelling or dilution with cheaper honeys or sugar syrups.
At present there is no single method for honey authenticity testing and existing methods are not readily available to the producers and distributors. Through-container Spatial Offset Raman Spectroscopy (SORS) is a through-container technique which can provide the chemical fingerprint of honeys and sugar syrups. In this study, SORS was employed on different types of natural honeys produced in the UK and Australia. Adulteration of honeys was simulated by mixing pure honeys with different types of sugar syrups from corn, sugar cane, sugar beet and rice at various concentrations as well as with cheaper imported honeys. The SORS data acquired were preprocessed and subjected to unsupervised multivariate analysis for pattern recognition purposes and to supervised analysis to construct prediction models for adulteration detection. Several statistical and machine learning algorithms were employed for this purpose, including PLS-DA, XGBoost and Random Forest to train classification and regression models. Random Forest consistently outperform the other algorithms in both classification and regression tasks and successfully differentiated between types of syrups. The results presented herein demonstrate the potential of SORS in combination with machine learning to be applied for the authentication of honey samples from different geographical areas and the detection of fraudulent practices such as exogenous sugar addition. The non-destructive nature, portability and speed of SORS makes it a suitable methodology for the construction of comprehensive honey authenticity databases and establish traceability across the supply chain.