Classification and geographic origin discrimination of Chinese prickly ash using X-ray fluorescence and machine learning

Chinese prickly ash (CPA) is a widely used seasoning with numb taste and a traditional Chinese medicine. Discrepancies among cultivating regions may pose a direct impact on the elemental contents, and ultimately on the flavor and quality of CPA. This study explored the elemental composition of CPAs by X-ray fluorescence spectrometry (XRF). Combined with chemometric methods such as principal component analysis, linear discriminant analysis, k-nearest neighbor, partial least squares discriminant analysis, support vector machine and random forest, the classification and discrimination of CPAs from different cultivation regions were realized with high accuracy. Taking advantage of XRF for direct non-destructive analysis, sample digestion essential in common analytical atomic spectrometry was omitted, which significantly simplified the sample pretreatment procedure and retained matrix components that are probably indicative of the geographic origins as well. The top ten important elements for discrimination were K, Sr, C, Ca, P, S, Fe, O, Mg and Si based on the importance score of random forest. The proposed strategy has been demonstrated as a simple, rapid yet efficient method for CPA classification, and may be applied to other food and medicinal samples.