Determination Geographical Origin and Flavonoids Content of Goji Berry Using Near-Infrared Spectroscopy and Chemometrics

Abstract

The feasibility of near-infrared (NIR) spectroscopy and chemometrics as tools to analyze Chinese Goji berry samples from four different topographical regions was investigated. Firstly, a consumer panel was asked to rate sensory attributes of the samples on a nine-point hedonic scale. Secondly, NIR original spectra of Goji berries in the wavelength range of 10,000–4000/cm were acquired. Least-squares support vector machine (LS-SVM) was firstly performed to calibrate the discrimination model to identify the geographical origins of the Goji berries, and the accuracy of correct identification was more than 96.67 %. Compared with artificial neural network (ANN) and K-nearest neighbors (KNN) approach, LS-SVM algorithm showed excellent generalization for identification results. Thirdly, as total flavonoid content (TFC) is highly related with the quality of the Goji berry, synergy interval partial least squares (Si-PLS) was applied to build the TFC prediction model. The determination coefficient for prediction (Rp) of the Si-PLS model was 0.9075, and root mean square error for prediction (RMSEP) was 0.376 mg/g. The three regions (4580–4860, 5720–6010, and 6290–6580/cm) selected by Si-PLS corresponded to the absorptions of two aromatic rings in the basic flavonoid structure. This work indicates that NIR spectroscopy combined with LS-SVM and Si-PLS offers significant potential and could be used as a rapid and efficient technique for evaluating the quality of retail Goji berries.