Discrimination of rice wine age using visible and near infrared spectroscopy combined with BP neural network

Visible and near infrared spectroscopy (Vis/NIR) combined with chemometric methods was employed to classify rice wines with different ages. Spectra of 240 wine samples (80 for each year) were collected in the Vis/NIR region (325-1075nm) in the spectroradiometer in transmission mode. Partial least squares (PLS) analysis was applied to extract the principal components (PCs) as new eigenvectors to represent the information of the raw spectra. Then the first five PCs were used as the inputs of the BP neural network. Finally, a four-layer BP neural networks model was developed 180 samples were selected randomly for the training set and the remaining 60 samples were for the prediction set. The threshold error of recognition was set as +/- 0.2. The discrimination ratio of 96.67% was achieved The results indicated that Vis/NIR spectroscopy could be used as a rapid alternative method to discriminate the rice wine age.