Discriminating Beef Producing Countries by Multi-Element Analysis and Machine Learning

The growing awareness of the environmental impact of beef production is greatly influencing the consumption decision. Beef production is strongly criticized due to the remarkable environmental impact of this activity, associated with problems of deforestation, water consumption, global warming, and climate change. Despite this, livestock food products play an important role in food security, accounting for 33% of global protein consumption. Enhancing the transparency of the beef production chain is essential to increase consumer perception about its origin, safety for consumption, environmental and human aspects. A study was undertaken to assess if beef samples from different producing countries can be distinguished from another on basis of their contents of chemical elements. Beef samples from some of the top world exporters, Brazil (1st), Australia (2nd), Argentina (5th), Uruguay (8th), and Paraguay (9th), were analyzed by neutron activation analysis for multi-element determination. Five machine learning algorithms, Classification and Regression Tree (CART), Multilayer Perceptron (MLP), Naive Bayes (NB), Random Forest (RF), and Sequential Minimal Optimization (SMO), were used to analyze the measurement results and classify the beef producing countries. MLP model provided the best classification performance, with an accuracy of 100%, 98%, 98%, 96%, and 82% respectively for Paraguay, Uruguay, Australia, Argentina, and Brazil. Reducing the number of classes (each country against the remaining countries), the accuracy achieved for the Brazilian beef samples was improved to 94% without changing the performance for other countries.Multi-element compositional data and machine learning algorithms allowed for discriminating beef producing countries, providing an outlook of becoming a valuable tool for geographical origin traceability and transparency