Rapid quality analysis of chinese rice wine by raman spectroscopy and support vector machines

In this study, Raman spectra analysis technology was used to rapidly and accurately evaluate various Chinese rice wines. Fingerprint information was collected from Chinese rice wine samples from different sources and of different ages, followed by a comparison of the prediction performances of quality evaluation models established by discriminant analysis (DA) and least-squares support vector machine (LS-SVM) in order to select the optimal model. Principal component analysis was used in the entire wavelength range to reduce the dimension of Raman spectroscopy data. The Mahalanobis distance between reduced-dimension spectra was calculated and discrimination analysis models were established based on ward's calculation. The entire wavelength range was used as the input of LS-SVM, to select a kernel radial basis function (RBF) that could alleviate non-linear issues. The kernel RBF was optimized at the same time via interactive authentication. The LS-SVM discrimination model was established based on the optimized kernel RBF function. The result showed that Raman spectroscopy coupled with LS-SVM could 100% accurately identify both production origin and wine age. The correction rates of Raman spectroscopy coupled with DA to discriminate Chinese rice wines from among Jiashan, Shaoxing, and Shanghai origins were 100%, 80%, and 80%, respectively, while that to determine wine age was 100% for all samples. Thus, the LX-SVM model was superior to the DA model. It was concluded that Raman spectroscopy in combination with chemometrics could be used for the rapid and accurate evaluation of Chinese rice wines. ©, 2015, South China University of Technology. All right reserved.