Time sequence marking method and device, equipment and storage medium

Abstract

The invention discloses a time sequence detection method and device, equipment and a storage medium. The method comprises the following steps: acquiring sequence points in a time sequence; obtaining afirst determination result of whether the sequence point is an abnormal point or not through a pre-constructed statistical model, and obtaining a second determination result of whether the sequence point is an abnormal point or not through a pre-constructed unsupervised learning model; if the first determination result is consistent with the second determination result, taking the sequence pointdetermined as a normal point as a normal sample, and taking the sequence point determined as an abnormal point as an abnormal sample; and obtaining a detection result of each sequence point in the time sequence through the classification model, and marking abnormal points in the time sequence according to the detection result. According to the technical scheme provided by the embodiment of the invention, the problems of missed detection and false detection when a single statistical model or an unsupervised learning model is adopted to detect the sequence points in the time sequence are avoided, and the accuracy and reliability of marking the abnormal points in the time sequence are improved.

Description

Technical field [0001] The embodiments of the present invention relate to the field of Internet technology, and in particular to a time series marking method, device, equipment, and storage medium. Background technique [0002] Time series refers to the ordered observation data set associated with time sequence obtained for a specific indicator in a certain application scenario. With the rapid development of Internet technology, it is necessary to predict and analyze the time series data corresponding to each indicator. Determine whether there are abnormal indicators in the time series. [0003] Most of the abnormal markers in the existing time series are manually detected by the engineer, or the time series is detected by a linear regression model to mark the corresponding abnormal points, but the engineer is required to have the corresponding application of the time series The business background of the scene and the large amount of sequence data that needs to be detected and la...

AI Technical Summary

Problems solved by technology

[0003] Most of the abnormal markers in the existing time series are manually detected by engineers, or anomaly detection is performed on the time series through a linear regression model to mark the corresponding abnormal points, but engineers are required to have the corresponding application of the time series The business background of the scene, and the large amount of sequence data that needs to be detected and labeled will consume a lot of labor costs; at the same time, the linear regression model has certain limitations and low real-time performance, and the reliability of the abnormal labeling results of the time series is not strong.

Images