NIR spectroscopy with grey wolf optimization algorithm for prediction of polyphenol content in inward tea leaves

Abstract

Polyphenol contributes significantly to the quality of tea. Non-invasive analysis of polyphenol in tea leaves is thus of prime importance. In this work, near infrared reflectance (NIR) and partial least squares (PLS) has been used to determine the total polyphenol content in tea leaves. During sample acquisition the number of variable is quite high for each spectra and whole range of spectra may not play an important role for building the calibration model of PLS algorithm. Thus, to determine the spectral region, Grey wolf optimizer (GWO) was used. Partial least squares (PLS) algorithm was used to generate the fitness function for GWO to estimate the total polyphenol content using the spectral region, found by optimization technique. During model calibration, for training and testing, leave-one-sample out cross-validation (LOSOCV) was used. The optimum range was obtained to be from 1043.5 nm to 1166 nm. The adequacy of the model developed was evaluated by root mean square error of cross-validation (RMSECV), root mean square error of prediction (RMSEP) and correlation coefficient (R). The RMSECV value found using GWO was 0.1364. The RMSEP and correlation coefficients (R) in the prediction set is 0.3244 and 0.91, respectively.