For the use of classifier in the structure damage diagnosis, a structural damage identification method is proposed based on wavelet packet decomposition and random forest which is a new combination classifier algorithm. Wavelet package decomposition is used to decompose the vibration acceleration signals of building structure with different damage degrees and locations. The energy sequences at different bands of frequencies decomposed by the wavelet packet decom position are inputted to classifier as feature vectors. A random forest model is trained and used to identify the location and degree of injury. The method is applied for damage identification of an eight-story shear steel frame model. Experimental results show that the method has good recognition accuracy and stability compared with BP neural network and support vector machines.