Dielectric relaxation parameters help to analyze protein content in cow's milk and improve prediction accuracy

Abstract

Can dielectric relaxation parameters aid the analysis of protein content in cow's milk and enhance the prediction accuracy? In this study, a modified Debye model was used to fit the dielectric spectra (20–4500 MHz) of 274 milk samples to obtain dielectric relaxation parameters (DRPs), and the dependence of DRPs on milk protein content was analyzed. Based on this investigation, the DRPs (εl, εh, Δε, τ, σ) were explored for the first time to be used for the prediction of protein content in cow's milk. The results showed that protein content mainly affected εl and Δε in the DRPs. The increase in protein content significantly inhibited the ionic polarization at low frequencies and weakened the orientational polarization of water molecules. The best performance came from the support vector regression model that combined DRPs with full spectra, with a root mean square error of prediction set (RMSEP) of 0.77 g kg-1 and a residual prediction deviation (RPD) of 4.447. The study demonstrated that the fusion of DRPs enhanced the prediction models based on full dielectric spectra. This method had higher prediction accuracy than previous research results without the need for sample or spectral preprocessing. The analysis of DRPs in this study contributes to understanding the effect mechanism of protein content on the dielectric properties of cow's milk. The proposed method of DRPs combined with dielectric spectra provides a new pathway to enhance the detection accuracy of cow's milk and other food components.