The invention relates to the technical field of nuclear radiation detection, and discloses a complex nuclide identification method based on a fuzzy decision tree. The method comprises the steps of performing smooth preprocessing on a radionuclide [gamma] energy spectrum by using a time-domain least squares method; performing peak-seeking on the processed [gamma] energy spectrum by using a symmetrical zero-area transformation method to extract a characteristic peak value; performing 5-layer one-dimensional wavelet transform on the processed [gamma] energy spectrum by using a db5 wavelet base pair hybrid energy spectrum to extract the maximum partial wave energy value after the transform; and integrating the characteristic information, ie, the characteristic peak value and the partial wave energy value to form a sample database, on the basis of a fuzzy decision tree method, dividing the characteristic space dynamically in the process of tree building, and constructing a nuclide identification classifier to realize nuclide identification. In this scheme, the partial wave energy after the wavelet transform is combined with the characteristic peak as the basis for the nuclide identification, and the reliability and stability of the identification are improved; and secondly, based on the fuzzy decision tree method in which the characteristic space is divided dynamically in the process of tree building, the identification accuracy of the decision tree is high.