Machine-learning assisted modelling of multiple elements for authenticating edible animal blood food

Elemental fingerprint coupled with machine learning modelling was proposed for species authentication of the edible animal blood gel (EABG). A total of 25 elements were determined by inductively coupled plasma mass spectrometry (ICP-MS) and atomic absorption spectroscopy (AAS) in 150 EABG samples prepared from five species of animals, namely duck, chicken, bovine, pig, and sheep. Extreme learning machine (ELM) models were constructed and optimized. Principal component analysis and Fisher linear discriminant analysis were comparatively utilized for dimension reduction of the crucial input elements selected via stepwise discriminant analysis and one-way ANOVA. The optimal ELM model was obtained with the crucial elements selected by one-way ANOVA from the relative content of the measured elements, which afforded accuracies of 98.0% and 96.0% for the training and test set, respectively. All findings suggest that elemental fingerprint accompanied by ELM have great potential in authenticating the edible animal blood food.