The method for identifying and classifying plastic express bags as the physical evidences commonly seen at crime scenes is proposed to promote the public security investigation. By means of X-ray fluorescence spectroscopy, the element types and contents of 60 plastic express bag samples with different origins and brands are tested. Meanwhile, the qualitative and semi-quantitative analysis is performed according to the spectral data, so that the 60 samples are preliminarily divided into 13 categories. Based on the within-cluster sum of squared errors (SSE), the relationship between SSE and number of clustersis explored, and the best number of clusters is determined to be 6. Besides, the 60 samples are clustered into 6 categories by using the K-means algorithm. Finally, the clustering results are analyzed through the Fisher discriminant analysis. As can be learned from the test, the accuracy of original classification and that after cross-validation classification of the 60 samples are 98.3% and 91.7%, respectively. This verifies that the method for identification and classification of plastic express bags is both accurate and scientific, which is based on K-means and SSE. The inspection and identification of classified plastic express bags as physical evidences through X-ray fluorescence spectroscopy, K-means and SSE are non-destructive, quick and effective, and the proposed model can give accurate and reliable results.