Xiaoting Yang, Shilin Zhao, Yuebo Sun, Chengqun Chen, Jing Liu, Di Wu, Yan Zhao*, Jie Cheng*

Food Chemistry；2026

Abstract:

Raman spectroscopy, with its rapid analytical advantages, shows great promise for food geographical origin traceability. The Stacking ensemble method improves generalization and accuracy by combining predictions from multiple heterogeneous base models. This study developed a beef geographical origin identification approach by integrating Raman spectroscopy with a Stacking ensemble model. Raman spectra were collected from beef samples across four regions. Seven beef origin identification models were constructed using five base models and two ensemble models (Stacking and Soft Voting). The results showed that the Stacking ensemble model outperformed individual base models and the Soft Voting ensemble model in all evaluation metrics, achieving perfect classification (precision, recall, F1, AUC, and Kappa all at 1.0) on both training and testing sets. Furthermore, the Stacking ensemble model maintained 92.5% accuracy in blind validation tests. This method provides a promising technical strategy for beef geographical origin authentication.