Shilin Zhao, Xiaoting Yang, Jing Liu, Guoyou Chen, Yalan Li, Yumeng Hu, Yan Zhao*, Baohai Liu *

Food Control；2026

Abstract:

Wuchang rice is the most famous premium rice of China and was included in the first batch of the China-EU Geographical Indications Agreement protection list. However, driven by economic interests, up to 90% of Wuchang rice on the market is counterfeit, originating from other regions. Therefore, it is imperative to establish effective methods for determining the origin of Wuchang rice. This study combined stable isotope (δ¹³C, δ¹⁵N, δ²H, and δ¹⁸O) with machine learning, including Bernoulli Naive Bayes (Bernoulli NB), Ridge Regression (RR), Stochastic Gradient Descent (SGD), Support Vector Machine (SVM), Random Forest (RF), and Extreme Gradient Boosting (XGBoost), to compare their performance and develop optimal models for distinguishing Wuchang rice from samples of other regions (Tailai, Jilin, Hubei, and Zhejiang). The results demonstrated that the XGBoost model outperformed other models, improving the classification accuracy from 90.20% with traditional chemometric methods to 97.14%. Furthermore, the XGBoost model achieved 100.00% discrimination accuracy for classifying two Wuchang rice cultivars, Wuyoudao 4 and Zhongkefa 5. These results provide a high-precision identification method for Wuchang rice.