Pet food categorization by neutron activation analysis and data science

Cat food samples from commercial brands available in the Brazilian market were characterized by neutron activation analysis. The multielemental profiles were evaluated according to animal age, feed category and main ingredient. Statistically significant differences were observed between feed category (standard, premium and super premium) and main ingredient (animal and vegetable). Machine learning algorithms were effective in discriminating feed category (78.6% accuracy) and main ingredient (78.7% accuracy), indicating that the multielemental profile can be used for identification and traceability of cat food. (author)