Non-invasive determination of enological parameters of rice wine by Vis-NIR spectroscopy and least-squares support vector machines

Abstract

The feasibility of visible and near-infrared (Vis-NIR) spectroscopy together with least-squares support vector machines (LS-SVM) for non-destructive determination of rice wine composition in coloured bottles was investigated. The Vis-NIR spectra were collected in a fiber spectrometer system with the bottle sealed and the labels removed. Statistical equations were established between the reference data and Vis-NIR spectra by partial least squares regression (PLSR) and LS-SVM, respectively. The prediction performance of LV-SVM is superior to that of PLSR, with higher rval of 0.960, 0.942 and 0.866, and lower RMSEP of 0.115 % (V V-1), 0.071 g L-1 and 0.021 for alcohol content, titratable acidity and pH, respectively. Based on the results, it was concluded that the Vis-NIR spectrometer system was suitable for non-destructive rice wine quality determination, and LS-SVM was a reliable multivariate method for rice wine composition prediction.