Determination of fat, protein and moisture in ricotta cheese by near infrared spectroscopy and multivariate calibration

Near-infrared reflectance spectroscopy (NIRS) with partial least squares regression (PLSR) was used to determine levels of fat, protein and moisture in ricotta cheese without complex sample preparation. Spectra of 19 conventional and low-fat ricotta samples from different manufacturers were collected in duplicate, with 33 of the 38 spectra used as a calibration set and the remaining 5 spectra used as an external validation set. The best results were obtained by processing the spectral region between 1,100 and 2,500 nm. Multivariate models with six latent variables (LVs) showed good prediction capability for fat and protein determinations, with average relative errors (Er) of 6.37 % and 5.95 %, respectively. For moisture, a more robust model was obtained with 4 LVs, showing better prevision capacity and Er of 1.91 %.