Application of near infrared spectroscopy and clustering analysis to classify wines from different origins

The quick, lossless and green near-infrared spectroscopy (NIRS) technology was explored to classify red wines according to their geographical origins namely Changli, Shacheng from China and Bordeaux from France. The NIRS preprocessed data collected from 47 red wine samples were analyzed by using stepwise regression analysis (SAR), principal components analysis (PCA) and cluster analysis (CA), and a prediction model of red wine geographical origin was established for discriminant analysis. The test result showed that the feature spectrum of the red wines of Changli, Shacheng and Bordeaux was 1 400–1 550 nm and 2 000–2 300 nm. The PCA space of red wines from three origins was basically independent distribution. It was the longest distance of spectra between the samples from Bordeaux and domestic origin. There was a little cross part between Changli and Shacheng samples. The clustering analysis model by the Ward’s method based on 38 samples was used to predict the 9 unknown samples. The origin recognition rate of 88. 9% was achieved. It is concluded that the NIRS technology is both accurate and stable for geographical origin wines discrimination.