Raman spectroscopy has been used in forensic science widely. In this paper, laser Raman spectroscopy analysis technology and K nearest neighbors algorithm are used to study 25 plastic steel window samples. The five principal components are extracted by principal component analysis, and the experiment built interactive verification test with the method regarding the training sample as the test sample. When the K value equals to 1, the lowest training sample error rate appears. Taking the three highest contribution value characteristic variables as parameters to build the classification model to realize the accurate classification of the unknown variables, and the total correct rate is 71%. The above method is more accurate than the direct observation of the spectra.