The ranchers handle donkeys breeding in small livestock farms and there is a need to supply a suitable tool to assess milk characteristics. To this aim, near infrared (NIR) spectroscopy was assessed through nine statistical methods: partial least squares, principal components, multivariate adaptive splines, regression trees using M5’ method, least squares support vector machine, artificial neural network, regression trees, ensemble regression trees and regularized least squares regression. These methods were implemented using a forward sequential feature selection algorithm; their predicting capability was validated with ten-fold cross validation; different types of spectra pre-processing methods were tested. The trials confirmed that NIR analysis is an efficient method for quantitative analysis of protein, lactose and dry matter content of donkey’s milk with a full-scale prediction error (FSPERR) of 3.0%, 4.4% and 4.5% respectively. Furthermore, a 16 NIR light emitting diodes (LED) device was investigated using simulation; in this case the prediction models of protein content showed the best FSPERR (3.1%), followed by those of dry matter content (7.0%) whereas those of the lactose content showed the worst (19.8%). These results could be used to design a low cost NIR LED device for the real-time control of donkey’s milk in breeding farms.

On-line measure of donkey’s milk properties by near infrared spectrometry

ALTIERI, Giuseppe;GENOVESE, FRANCESCO;ADMANE, NAOUEL;DI RENZO, Giovanni Carlo
2016

Abstract

The ranchers handle donkeys breeding in small livestock farms and there is a need to supply a suitable tool to assess milk characteristics. To this aim, near infrared (NIR) spectroscopy was assessed through nine statistical methods: partial least squares, principal components, multivariate adaptive splines, regression trees using M5’ method, least squares support vector machine, artificial neural network, regression trees, ensemble regression trees and regularized least squares regression. These methods were implemented using a forward sequential feature selection algorithm; their predicting capability was validated with ten-fold cross validation; different types of spectra pre-processing methods were tested. The trials confirmed that NIR analysis is an efficient method for quantitative analysis of protein, lactose and dry matter content of donkey’s milk with a full-scale prediction error (FSPERR) of 3.0%, 4.4% and 4.5% respectively. Furthermore, a 16 NIR light emitting diodes (LED) device was investigated using simulation; in this case the prediction models of protein content showed the best FSPERR (3.1%), followed by those of dry matter content (7.0%) whereas those of the lactose content showed the worst (19.8%). These results could be used to design a low cost NIR LED device for the real-time control of donkey’s milk in breeding farms.
File in questo prodotto:
Non ci sono file associati a questo prodotto.

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: http://hdl.handle.net/11563/115226
 Attenzione

Attenzione! I dati visualizzati non sono stati sottoposti a validazione da parte dell'ateneo

Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus 10
  • ???jsp.display-item.citation.isi??? 7
social impact