Application of visible and near-infrared spectroscopy for evaluation of ewes milk with different feeds A. Bahri A,B , S. Nawar C,E , H. Selmi D , M. Amraoui A , H. Rouissi A and A. M. Mouazen C,E A Department of Animal Production, Higher School of Agriculture of Mateur, 7030 Mateur, Tunisia. B National Agronomy Institute Tunis, 43 Avenue Charles Nicolle, Tunis 1082. C Department of Environment, Ghent University, Coupure links 653, 9000 Gent, Belgium. D Sylvo-Pastoral Institute of Tabarka, University of Jandouba, BP.n  345, Tabarka 8110, Tunisia. E Corresponding author. Email: said.nawar@ugent.be; abdul.mouazen@ugent.be Abstract. Rapid measurement optical techniques have the advantage over traditional methods of being faster and non-destructive. In this work visible and near-infrared spectroscopy (vis-NIRS) was used to investigate differences between measured values of key milk properties (e.g. fat, protein and lactose) in 30 samples of ewes milk according to three feed systems; faba beans, field peas and control diet. A mobile fibre-optic vis-NIR spectrophotometer (350–2500 nm) was used to collect reflectance spectra from milk samples. Principal component analysis was used to explore differences between milk samples according to the feed supplied, and a partial least-squares regression and random forest regression were adopted to develop calibration models for the prediction of milk properties. Results of the principal component analysis showed clear separation between the three groups of milk samples according to the diet of the ewes throughout the lactation period. Milk fat, protein and lactose were predicted with good accuracy by means of partial least-squares regression (R 2 = 0.70–0.83 and ratio of prediction deviation, which is the ratio of standard deviation to root mean square error of prediction = 1.85–2.44). However, the best prediction results were obtained with random forest regression models (R 2 = 0.86–0.90; ratio of prediction deviation = 2.73–3.26). The adoption of the vis-NIRS coupled with multivariate modelling tools can be recommended for exploring to differences between milk samples according to different feed systems, and to predict key milk properties, based particularly on the random forest regression modelling technique. Additional keywords: chemometrics, feeding, random forest. Received 19 April 2017, accepted 29 May 2018, published online 1 August 2018 Introduction Ewes’ milk is considered to have a high nutritional value with high concentrations of proteins, fats, minerals, and vitamin and is commonly used for chees making (Balthazar et al. 2017). Ewes diet is the primary factor for maintaining milk quality. Sicilo- Sarde ewes are the most important dairy sheep breed in Tunisia. Their main sources of feed are silage, rangeland hay, stubble and crop residues, which have low nutritional value and did not cover the nutritional requirement of ewes. This is the reason why these feeds are usually complemented with concentrate feed (Rouissi et al. 2008). The concentrate feed is formed mainly of imported raw materials, for example, corn as an energy source and soybeans as a protein source, which weigh down the national economy and prevent the development of ewe’s milk production. Rouissi et al.(2008) and Hammami et al.(2013) searched for other feeds cultivable in the Tunisian farms, which would not detrimentally affect milk quality. Changing the concentrate feed affects the quality of milk, whose properties need to be measured rapidly and cost effectively using a portable system that can be taken to the farm. Laboratory reference analytical methods for milk use expensive and polluting reactive agents and relay on qualified technicians. In addition, they are time consuming and destructive methods. Other methods, such as those based on ultrasonic analysis may provide information on the physical and chemical properties of foods, including composition and physical state of milk products (Knorr et al. 2004; Leadley and Williams 2006). However, the use of ultrasonic waves in milk may cause loss of the proteolytic activity and the milk becomes less coagulable (Raharintsoa et al. 1978). Optical methods including the visible and near-infrared red spectroscopy (vis-NIRS) has been widely used as alternative measurement method in different research areas of research, including but not restricted to dairy products, soil, feed, food, manure, and archaeology. This technique is simple, sensitive, non-destructive, rapid and can be used on-site to collect high density information at low cost and allow several parameters to be monitored simultaneously (Birlouez-Aragon et al. 2002; Kulmyrzaev et al. 2008). Vis-NIRS has been widely used in the milk industry, for quantitative and qualitative analyses in raw milk, cheese and CSIRO PUBLISHING Animal Production Science https://doi.org/10.1071/AN17240 Journal compilation Ó CSIRO 2018 www.publish.csiro.au/journals/an