Quality determination of fruit vinegars using visible/near infrared spectroscopy and least squares-support vector machine

Soluble solids content (SSC) and pH are two important quality parameters of fruit vinegars. Visible and near infrared (VIS/NIR) spectroscopy was employed to determine SSC and pH of fruit vinegars based on partial least squares (PLS) analysis and least squares-support vector machine (LS-SVM). 300 vinegar samples were prepared. PLS models were developed with different preprocessing methods including no treatment, smoothing way of Savitzky-Golay, standard normal variate, 1st- and 2nd-derivertives. Simultaneously, certain selected latent variables (LVs) were used as LS-SVM inputs according to their explained variance. Finally, LS-SVM models with RBF kernel were developed compared with PLS models. The raw spectral data showed the best performance. The best LS-SVM models were achieved with 7 LVs for SSC and 6 LVs for pH, and LS-SVM outperformed all PLS models. The correlation coefficient (r), RMSEP and bias for validation set were 0.980, 0.667 and -0.043 for SSC, whereas 0.992, 0.040 and -0.006 for pH, respectively. The overall results indicated that VIS/NIR spectroscopy combined with LS-SVM could be used as a rapid alternative method for the prediction of SSC and pH of fruit vinegars. These results would be helpful for the process monitoring during the fermentation of fruit vinegars.   Key word: Visible and near infrared spectroscopy, fruit vinegar, soluble solids content, pH, partial least squares analysis, least squares-support vector machine.