Perturbation theory for cross data matrix-based PCA

Discrimination and quantification of volatile compounds in beer by FTIR combined with machine learning approaches

The composition of volatile compounds in beer is crucial to the quality of beer. Herein, we identified 23 volatile compounds, namely, 12 esters, 4 alcohols, 5 acids, and 2 phenols, in nine different beer types using GC-MS. By performing PCA of the data of the flavor compounds, the different beer types were well discriminated. Ethyl caproate, ethyl caprylate, and phenylethyl alcohol were identified as the crucial volatile compounds to discriminate different beers. PLS regression analysis was performed to model and predict the contents of six crucial volatile compounds in the beer samples based on the characteristic wavelength of the FTIR spectrum. The R2 value of each sample in the prediction model was 0.9398-0.9994, and RMSEP was 0.0122-0.7011. The method proposed in this paper has been applied to determine flavor compounds in beer samples with good consistency compared with GC-MS.