Non-Linear Modeling of Water Content in Processed Green Tea Using ANN and PCA

Application of the Hyphenated PCA-ANN regression Model to the Absorbance Obtained by NIRS

Abstract

For determination of moisture of the processed Japanese green tea, Principal Component Analysis (PCA) and Artificial Neural Networks (ANN) were applied to the absorbance that were provided with near infrared spectoroscopy (NIRS). The 3-3-1 hierarchy type PCA-ANN (principal component analysis-artificial neural networks) regression model used the first three PCA scores as input signal, and the water content as output signal, showed that the standard error of prediction (SEP) and the correlation coefficient (r) between the actual value and the predicted value of water were 1.547%w. b. and 0.998 in prediction set respectively. This value of SEP was decreasing to about 30-73%, compared to the conventional multiple linear regression (MLR), principal component regression (PCR) and partial least squares (PLS) regression model, and this model performed with a high accuracy of prediction in prediction set. The efficiency of the combined model of PCA-ANN regression becomes clear as the modeling for multicollinearity in variables and non-linearity between the moisture content covering a wide range in tea processing and the absorbance provided with NIRS.