The invention relates to an optical fiber vibration signal feature extraction and classification method based on DWT-DFPA-GBDT, is a method for feature extraction and classification of optical fiber vibration signals, and belongs to the field of signal processing and machine learning. The optical fiber vibration signal feature extraction and classification method is characterized by comprising thefollowing steps: (1) determining time domain features of original vibration signals; (2) determining frequency domain characteristics of the original vibration signal; (3) determining wavelet domaincharacteristics of the original vibration signal; (4) calculating a Mahalanobis distance; (5) obtaining the sorting of the feature vectors according to the separability; (6) determining a loss function; (7) determining a fitting function; and (8) determining a strong learning classifier of the current iteration. According to the optical fiber vibration signal feature extraction and classificationmethod, feature extraction of three different domains of a time domain, a frequency domain and a wavelet domain is realized; a complete vibration signal feature vector is constructed; and GBDT and DWTare combined, and a method for reducing the complexity of a model on the basis of ensuring the classification precision is provided for the field of vibration signal feature extraction and classification.