Prediction of fatty acid content in sheep milk by Mid-Infrared spectrometry with a selection of wavelengths by Genetic Algorithms

Abstract

Sheep breeding is one of the most widespread activities in Sardinia (Italy), and milk produced here is of crucial economic importance for the region. In order to make the milk payment system used in Sardinia more rewarding to the quality of milk, we developed Partial Least Square regression models to predict the concentration of the major fatty acids (measured with a GC-FID reference method) from the Mid-Infrared spectra of hundreds of Sardinian sheep milk samples collected in the period 2011–2013. Genetic Algorithms were used in order to select the most informative spectral subsets and therefore reduce the complexity of the model and in many cases also reduce the prediction error. Models obtained had a good predictive ability, with errors in the range of tenths of a gram of fatty acid on Kg of milk, and an acceptable precision for an immediate introduction on sheep milk payment in Sardinia.