Beef Traceability Between China and Argentina Based on Various Machine Learning Models

Area of Science:

• Food Science
• Analytical Chemistry
• Computational Biology

Background:

• Consumer demand for high-quality beef necessitates reliable origin tracing.
• Current beef origin identification methods require enhancement for accuracy and efficiency.
• Production region significantly impacts beef's nutritional value and quality.

Purpose of the Study:

• To develop a robust classification model for predicting beef origin.
• To analyze elemental content and stable isotopes for origin determination.
• To compare the performance of different machine learning algorithms for beef traceability.

Main Methods:

• Elemental analysis of 52 elements using ICP-MS and ICP-OES.
• Stable carbon isotope ratio determination via EA-IRMS.
• Construction and evaluation of machine learning models including PLS-DA, CNN, and Random Forest.

Main Results:

• PLS-DA model achieved 98.8% classification accuracy and 94.12% prediction accuracy.
• Key elements (Fe, Cs, As, Co, V, Sc, Rb, Ru) and δ13C were identified as crucial for origin prediction.
• PLS-DA model showed superior performance with R² of 0.924 and Q² of 0.787.

Conclusions:

• Combining elemental and stable isotope analysis with machine learning effectively traces beef origin.
• The developed models enhance food safety and meet consumer demand for verified beef provenance.
• This approach provides a reliable method for differentiating beef from various production regions.