Back Propagation-Artificial Neural Network Model for Prediction of the Quality of Tea Shoots through Selection of Relevant Near Infrared Spectral Data via Synergy Interval Partial Least Squares

Near-infrared spectroscopy and back propagation-artificial neural network (BP-ANN) model in conjunction with synergy interval partial least squares (siPLS) algorithm were used to evaluate tea shoots quality. The near-infrared spectra regions relevant to tea quality (12493 cm-1 to 11645 cm-1, 9087.5 cm-1 to 8242.7 cm-1, 8238.9 cm-1 to 7394.2 cm-1, and 6541.7 cm-1 to 5697 cm-1) were selected using siPLS algorithm. The two principal components that explained 99.46% of the variability in this spectral data were then used to calibrate the BP-ANN quality index (QI) model. The performance of this model [the coefficient of determination for prediction (), 0.9680; root mean square error of prediction (RMSEP), 0.0178] was superior to those of the BP-ANN model ( = 0.9332, RMSEP = 0.0285) and the siPLS model ( = 0.9230, RMSEP = 0.0360). The predicted QI values of 25 samples highly correlated with the experimental values ( = 0.9223, RMSEP = 0.0344). The QI model with the combined siPLS-BP-ANN algorithms accurately predicted the quality of tea shoots.