Authentication of beef cuts by multielement and machine learning approaches

Area of Science:

• Food Science
• Analytical Chemistry
• Computational Biology

Background:

• Brazil is a major global beef producer and exporter.
• Beef's multielement profile is influenced by genetics, diet, origin, and climate.
• Accurate authentication of beef cuts is crucial for market integrity.

Purpose of the Study:

• To evaluate the use of major and trace elements in beef authentication.
• To assess the effectiveness of various classification algorithms for beef cut identification.
• To determine the optimal algorithm for accurate beef traceability.

Main Methods:

• Neutron activation analysis (NAA) was used to determine major and trace elements (Br, Co, Cs, Fe, K, Na, Rb, Se, Zn) in beef samples.
• Supervised learning algorithms including Classification and Regression Tree (CART), Multilayer Perceptron (MLP), Naïve Bayes (NB), Random Forest (RF), and Sequential Minimal Optimization (SMO) were employed.
• Classification models were built using the multielement profiles of chuck steak, rump cap, and sirloin steak from Angus, Nelore, and Wagyu crossbreeds.

Main Results:

• The multielement profiles of beef cuts were successfully determined.
• Classification accuracies varied among algorithms: MLP (96%), SMO (95%), RF (91%), NB (86%), and CART (70%).
• The Multilayer Perceptron (MLP) algorithm demonstrated the highest classification performance.

Conclusions:

• Multielement profiling combined with machine learning algorithms is a viable tool for beef cut authentication.
• The Multilayer Perceptron (MLP) algorithm offers the best performance for identifying beef cuts based on elemental composition.
• This approach enhances food traceability and ensures the integrity of beef products.