Abstract-Generalized regression neural network association with terahertz spectroscopy for quantitative analysis of benzoic acid additive in wheat flour

Xudong Sun, Junbin Liu, Ke Zhu, Jun Hu,  Xiaogang Jiang, Yande Liu

https://royalsocietypublishing.org/doi/full/10.1098/rsos.190485

Investigations were initiated to develop terahertz (THz) techniques associated with machine learning methods of generalized regression neural network (GRNN) and back-propagation neural network (BPNN) to rapidly measure benzoic acid (BA) content in wheat flour. The absorption coefficient exhibited a maximum absorption peak at 1.94 THz, which generally increased with the content of BA additive. THz spectra were transformed into orthogonal principal component analysis (PCA) scores as the input vectors of GRNN and BPNN models. The best GRNN model was achieved with three PCA scores and spread value of 0.2. Compared with the BPNN model, GRNN model to powder samples could be considered very successful for quality control of wheat flour with a correlation coefficient of prediction (rp) of 0.85 and root mean square error of prediction of 0.10%. The results suggest that THz technique association with GRNN has a significant potential to quantitatively analyse BA additive in wheat flour.

Abstract-Terahertz Spectroscopy Determination of Benzoic Acid Additive in Wheat Flour by Machine Learning

Xudong Sun, Ke Zhu, Junbin Liu, Jun Hu, Xiaogang Jiang, Yande Liu Zhiyuan Gong

https://link.springer.com/article/10.1007%2Fs10762-019-00579-z

Investigations were carried out to develop terahertz (THz) techniques combined with machine learning methods for rapidly measuring benzoic acid (BA) additive in wheat flour. The absorption coefficient had a significant characteristic peak at 1.94 THz, which usually varies with the increase of BA additive concentrations. Correlation analysis was initiated to choose a frequency pair of 1.94 and 1.86 THz that improved the predictive ability of multiple linear regressions (MLR) model. Then, quantitative determination models were modeled between BA concentrations and THz absorption coefficients using partial least square (PLS) regression and least square support vector machine (LS-SVM). Compared with MLR and PLS models, the LS-SVM model could be regarded as an effective tool for quality control of wheat flour with a correlation coefficient of prediction (Rp) of 0.994 and root mean square error of prediction (RMSEP) of 0.12%. The results suggest that combining THz spectroscopy with LS-SVM has the potential to quantitatively analyze BA additive in wheat flour.