Abstract:

A high performance liquid chromatographic method was developed to establish the fingerprint of wines, and 36 samples from various manufacturers and various batches in Jiangxi province were analyzed. In this study, the technique of projection discriminance based on principal component analysis (PCA) and cluster analysis (CA) was used to differentiate and evaluate the fingerprints, and then, PCA was also employed to handle the data of the common chromatographic fingerprints pattern to reduce the number of variables, thus optimizing back-propagation network (BPN), which was applied to predict the attribution of unknown samples. The results on PCA and CA showed that there are definite differences among the wine samples produced by different manufacturers, based on which a method can be established to distinguish wine samples produced by different manufacturers, and the developed method can provide some scientific basis for the quality control of wines. And then, PCA was secondly adopted to optimize back-propagation network. The multiple predicted results manifested that the optimized network has high accuracy and good stability. PCA-BPN technology can be used to predict correctly the attribution of unknown samples.

Key words: high performance liquid chromatography, fingerprints, wines, principal component analysis, cluster analysis, back-propagation network