The work aims to explore a test method based on spectral analysis to achieve rapid and accurate identification of cigarette case evidence. Thirty-nine different yellow cigarette cases were tested with portable differential Raman spectrometer. The differential Raman spectra data of each sample were obtained. The samples were preliminary classified according to the type of filler. On the basis of data dimension reduction through principal component analysis, the results were clustered by Hierarchical clustering and K-Means clustering with IBM SPSS Statistics 26.0 software. The 39 samples could be divided into 3 categories according to the type of filler. And they could be further divided into 6 categories in combination with chemometrics. The results show that the method is non-destructive, fast and accurate, and the spectrum is not affected by fluorescence interference. Combined with chemometrics method, the cigarette case samples can be classified and tested. This method provides a basis for public security organs to test such material evidence at the scene of crime.