Identification of Poria Origin Based on Multi-Matrix Projection Discrimination of PCA

This study proposes a rapid method for identifying the geographical origin of Poria by combining Raman spectroscopy with an improved PCA algorithm—multi-matrix projection discrimination analysis. Poria samples from four Chinese provinces—Yunnan, Anhui, Shaanxi, and Hubei—were analyzed. Four datasets were constructed, each containing 25 Raman spectra per origin, with an additional 10 spectra per origin reserved as independent test sets. PCA was then separately applied to the spectral dataset of each origin to derive its respective eigenvector matrix. For each test spectrum, four reconstructed spectra were generated by projecting it onto the eigenvector matrices of the four origins. The origin was determined by identifying the one with the minimum Euclidean distance between the test spectrum and its reconstructions. When the first six principal components were used for model construction, the test set accuracy reached 97.5%, significantly outperforming the optimized PCA–SVM model, which achieved an accuracy of 85%. These results demonstrate that Raman spectroscopy, combined with the multi-matrix projection discrimination method based on PCA, can effectively capture the fingerprint information of Poria and accurately determine its geographical origin.