Robust model of fresh jujube soluble solids content with near-infrared (NIR) spectroscopy
A robust partial least square (PLS) calibration model with high accuracy and stability was established for the measurement of soluble solids content (SSC) of fresh jujube using near-infrared (NIR) spectroscopy
technique. Fresh jujube samples were collected in different areas of Taigu and Taiyuan cities, central China in 2008 and 2009. A partial least squares (PLS) calibration model was established based on the NIR spectra of 70 fresh jujube samples collected in 2008. A good calibration result was obtained with correlation coefficient (Rc) of 0.9530 and the root mean square error of calibration (RMSEC) of 0.3951 °Brix. Another PLS calibration model was established based on the NIR spectral of 180 samples collected in 2009; it resulted in the Rc of 0.8536 and the RMSEC of 1.1410 °Brix. It could be seen that the accuracy of established PLS models were different when samples harvested in different years were used for the model calibration. In order to improve the accuracy and robustness of model, different numbers (5, 10, 15, 20, 30 and 40) of samples harvested in 2008 were added to the calibration sample set of the model with samples harvested in 2009, respectively. The established PLS models obtained Rc with the range of 0.8846 to 0.8893 and RMSEC with the range of 1.0248 to 0.9645 °Brix. The obtained results were
better than the result of the model which was established only with samples harvested in 2009. Moreover, the models established using different numbers of added samples had similar results. Therefore, it was concluded that adding samples from another harvest year could improve the accuracy and robustness of the model for SSC prediction of fresh jujube. The overall results proved that the consideration of samples from different harvest places and years would be useful for establishing an accuracy and robustness spectral model.
Keywords: Near-infrared (NIR) spectroscopy, Huping jujube, soluble solids content (SSC), partial least squares (PLS), accuracy, stability