Honey Botanical Origin Authentication Using HS-SPME-GC-MS Volatile Profiling and Advanced Machine Learning Models...

Area of Science:

• Analytical Chemistry
• Food Science
• Cheminformatics

Background:

• Honey botanical authentication is crucial for quality control and fraud detection.
• Traditional methods often lack the sensitivity and specificity for complex mixtures.
• Volatile organic compound (VOC) profiling offers a promising avenue for discrimination.

Purpose of the Study:

• To develop and compare advanced machine learning models for multiclass honey botanical origin authentication.
• To integrate Headspace Solid-Phase Microextraction Gas Chromatography-Mass Spectrometry (HS-SPME-GC-MS) with supervised learning algorithms.
• To identify key chemotaxonomic markers for differentiating honey floral sources.

Main Methods:

• Analysis of 57 honey samples using HS-SPME-GC-MS to obtain VOC profiles.
• Application of Principal Component Analysis (PCA) for data visualization and class separation.
• Training and evaluation of Random Forest (RF), eXtreme Gradient Boosting (XGBoost), and Neural Network (NN) models for classification.

Main Results:

• Identification of specific VOC markers for coriander, orange blossom, astragalus, rosemary, and chehelgiah honeys.
• PCA demonstrated clear separation of botanical classes.
• The Neural Network model achieved the highest authentication accuracy (90.32%), outperforming RF and XGBoost.

Conclusions:

• HS-SPME-GC-MS combined with deep learning models provides a rapid, sensitive, and reliable method for honey botanical authentication.
• This approach offers significant potential for real-time quality control and combating honey adulteration.
• The study highlights the efficacy of advanced machine learning in complex food analysis.