A Hybrid Variable Selection Strategy of Near Infrared Spectroscopy for Detection the Ratio of Tea Polyphenols to Amino Acids in Green Tea Infusion

Abstract

Objective To establish the green and fast method of near infrared spectroscopy（NIR）for the determination of flavor quality index of the ratio of tea polyphenols to amino acids in green tea infusion，the chemometric method was used to improve the detection accuracy. Method First，standard normal variate transformation（SNV）was employed to preprocess the original NIR spectra of tea infusion to eliminate the noise caused by light scattering and improve the spectral signal-to-noise ratio.Then，a hybrid variable selection strategy based on variable combination population analysis and iteratively retains informative variables（VCPA-IRIV），competitive adaptive reweighted sampling and successive projections algorithm were used to optimize the wavelength variable.Using the optimization feature variables，a nonlinear random forest（RF）prediction model was built and compared it with linear partial least squares（PLS）. Result The results showed that the 18 characteristic wavelengths selected by VCPA-IRIV were related to tea polyphenols and amino acids，and avoid the interference of the strong correlation spectrum characteristic information of the tea soup water.The results of the models indicated that the performance of the RF model was better than that of PLS model.The performance of the RF model obtained was as follows：the correlation coefficient（Rc）and root mean square error（RMSEC）of calibration set were 0.949 and 0.231，respectively；the correlation coefficient（Rp），root mean square error（RMSEP）and relative percent deviation（RPD）of prediction set were 0.943，0.232 and 3.019，respectively. Conclusion The overall results demonstrated that the VCPA-IRIV method could effectively extract effective NIR spectral variables，eliminate redundant spectral information.The study provides new research ideas for rapid detection methods for determination of agricultural product quality and safety by means of near-infrared spectroscopy.