Abstract: This study established a new method for the classification and identification of livestock and poultry meat by integrating high-resolution inductively coupled plasma-tandem mass spectrometry (HR-ICP-MS/MS) and neural network models. The concentrations of multiple elements in different types of livestock and poultry meat were determined using HR-ICP-MS/MS, followed by visualization of the characteristic distribution of elements. This study further analyzed correlation patterns among different elements and developed neural network models. The results showed that the HR-ICP-MS/MS method exhibited excellent linearity for quantifying 49 elements in meat, with a coefficient of determination (R2) ≥ 0.996 6, and its detection limit and precision were superior to those of the method specified in the current national food safety standard. The visualization of elemental distribution revealed that different types of animal-derived products exhibited distinct elemental correlation patterns and significant biomarker heterogeneity. A random forest model was employed to accommodate high-dimensional features. After training, the model achieved a classification accuracy of 97.5% on the test set, with a recall of 97.7%, a precision of 98.2%, and an area under the receiver operating characteristic curve of 0.997. The model had good balance and could enable the classification and traceability of different types of livestock and poultry meat.

Key words: livestock and poultry meat; inductively coupled plasma-tandem mass spectrometry; random forest; classification and traceability