Excitation-emission matrix fluorescence spectroscopy combined with chemometrics methods for rapid identification and quantification of adulteration in Atractylodes macrocephala Koidz

• Fluorescence EEMs combined with chemometrics method was proposed to identify the Atractylodes macrocephala Koidz. adulterated with different grain powders. • Classification model for adulteration identification was built by PLS-DA, k NN and random forest. • All the classification models achieved satisfactory results for the holdout adulterated AMK samples. • Adulteration levels in AMK can be predicted well by PLS regression model. Atractylodes macrocephala Koidz. (AMK) is a perennial herb with various medical functions and has been wildly used in ethno-medical system. It is common that unscrupulous merchants try to make huge profits by adulterating AMK powder with other cheaper or lower quality edible powder substance due to the growing shortage of genuine medicinal materials resources and the rising cost. Therefore, this work proposed excitation-emission matrix fluorescence spectroscopy combined with chemometrics methods for the rapid identification and quantification of AMK adulteration with other types of powder. Partial least squares discriminant analysis (PLS-DA), k -nearest neighbor ( k NN) and random forest (RF) model were used for the classification of pure AMK and specific type powder adulterated AMK. The correct classification rates for test sample were 93.0%, 95.0% and 100% for k NN, PLS-DA and RF, respectively. And RF could accurately classify 11 holdout adulterated AMK samples, even when the adulteration level was only 10%. Furthermore, the PLS regression model was used for the prediction of adulteration level in AMK. The results proved that the classification and regression models were reliable.