Deep learning-augmented inductively coupled plasma atomic emission spectrometry for multivariate authentication of...

The origin and grade of green tea significantly influence its market value, yet their concurrent authentication remains challenging. Here, we developed a robust method combining multiple metallic elements and back propagation neural network (BPNN) to identify the origin and grade of tea simultaneously. This strategy utilizes inductively coupled plasma atomic emission spectrometry (ICP-AES) to analyze the content of multiple elements in green tea (e.g. Longjing tea) samples. BPNN achieved identification accuracies of 92.59 % for the origins and 88.89 % for the grades. Subsequently, Shapley Additive Explanations (SHAP) was used for visualizing the model's identification process and comparing the dimensionality reduction performance with principal component analysis (PCA). Although both methods showed similar performance in identifying origins, SHAP-based optimization proved more effective in classifying tea grades. The strategy proposed in this paper establishes a simple and reliable method for the simultaneous identification and authentication of the origin and grade of tea.