The invention relates to the field of the Raman spectrum detection, and specifically relates to a Raman spectrum data de-noising method based on a five-spot cubic smoothing algorithm. The method comprises the following steps: (1) performing base line removal processing on an acquired actual measurement sample Raman spectrum to obtain an initial spectrum; (2) performing five-spot cubic smoothing onthe initial spectrum obtained in the step (1); (3) performing m-1 times of five-spot cubic smoothing in the step (2) by taking the spectrum data obtained in the step (2) as the initial spectrum, thereby obtaining a final output spectrum, wherein the m is number of iterations, and m is equal to 3 to 5. The random interference error of the Raman spectrum data can be effectively filtered, and the characteristic feature of a to-be-measured component signal is highlighted, thereby ensuring the smoothness of the spectrum data; the signal-to-noise ratio of an original spectrum signal is improved, and the analysis accuracy of the spectrum data can be effectively improved.