Raman spectroscopy combined with stacking ensemble modeling for tracing the geographical origin of beef

Area of Science:

• Food science
• Analytical chemistry
• Machine learning

Background:

• Geographical origin traceability is crucial for food safety and authenticity.
• Raman spectroscopy offers rapid and non-destructive analysis for food authentication.
• Ensemble methods can enhance the accuracy of classification models.

Purpose of the Study:

• To develop a beef geographical origin identification approach using Raman spectroscopy and a Stacking ensemble model.
• To evaluate the performance of the Stacking ensemble model against individual base models and a Soft Voting ensemble.
• To assess the model's effectiveness in real-world blind validation scenarios.

Main Methods:

• Collected Raman spectra from beef samples originating from four distinct geographical regions.
• Constructed seven beef origin identification models, including five base models and two ensemble models (Stacking and Soft Voting).
• Evaluated model performance using metrics such as precision, recall, F1-score, AUC, and Kappa.

Main Results:

• The Stacking ensemble model achieved perfect classification (1.0 for all metrics) on both training and testing datasets.
• The Stacking model significantly outperformed individual base models and the Soft Voting ensemble.
• Achieved 92.5% accuracy in blind validation tests, demonstrating robustness.

Conclusions:

• The integration of Raman spectroscopy and the Stacking ensemble model provides a highly accurate and reliable method for beef geographical origin authentication.
• This approach offers a promising technical strategy for enhancing food traceability systems.
• The Stacking ensemble method demonstrates superior generalization and accuracy for complex classification tasks in food analysis.