Quasi-periodic time sequence unsupervised anomaly detection method, system and terminal

The invention discloses a quasi-periodic time sequence unsupervised anomaly detection method, system and terminal, and belongs to the technical field of time sequence anomaly detection, and the method comprises the steps: carrying out the segmentation of a quasi-periodic time sequence; performing clustering processing on the quasi-periodic time subsequences to obtain a data set formed by normal quasi-periodic time subsequences; and training the neural network model based on the data set to obtain an unsupervised anomaly detection model. According to the method, the quasi-periodic time sequence is firstly segmented, so that the quasi-periodic time subsequence is used as a detection object in the subsequent anomaly detection process, and the subsequent anomaly detection result has interpretability; and meanwhile, preprocessing is performed through clustering, and then the model is trained through the normal quasi-periodic time subsequences, so that the model learns data distribution of the normal quasi-periodic