X-ray fluorescence spectrometry is employed to conduct three tests on each of 80 blade samples, resulting in a total of 240-set spectral data. After preprocessing, feature elements are selected based on the ratio of the relative standard deviation of elements among samples to the mean relative standard deviation from three tests. These chosen feature elements included Fe, Cr, Mn, Cu, Ni, Ti, Pb, Ca, Mo, Zn, Ga, and Nb. Subsequently, data for 12 feature elements are subjected to Z-score standardization to eliminate dimensional differences among elements. Visual analysis and principal component analysis are then performed. Finally, a Bayesian-optimized random forest algorithm is employed for the classification and identification of these 80 samples, and it achieves an accuracy rate of 95%. Cross-validation results in an average accuracy of 92.5% with a standard deviation of 1.02%. Results of this research demonstrate that the combination of X-ray fluorescence spectrometry and the random forest algorithm can effectively achieve sample identification, provid a method by which to trace the brands and series of blade evidence from crime scenes, and offer valuable leads for investigative purposes.