Time series anomaly detection method and system for key performance index data

Abstract

The invention provides a time series anomaly detection method for key performance index data, which is used for solving the problems of low time series data anomaly detection efficiency and low accuracy in the prior art. The a time series anomaly detection method comprises the following steps: correcting missing values and abnormal values in acquired time series data; extracting features to splicethe data; dividing the spliced data into a training set and a test set; performing training on the basis of optimizing a target function to obtain an anomaly detection model; performing testing and evaluating to obtain an evaluation standard; and performing time series anomaly detection on the data to be detected according to the evaluation standard and the anomaly detection model. According to the invention, anomaly detection is carried out on different time series data with periodicity; features of different dimensions are extracted from the time series data to ensure that the correlation of the data in different dimensions can be learned by the model, so that the cost caused by abnormal labeling is reduced, and meanwhile, the method is suitable for a scene with extremely non-uniform positive and negative samples, and the detection efficiency is improved.

Description

technical field [0001] The invention belongs to the field of data processing and security, and in particular relates to a timing anomaly detection method and system for key performance index data. Background technique [0002] With the rapid development of data collection and storage technology, a large amount of time series data has been accumulated in the fields of finance, transportation, and the Internet. In order to ensure that various services in the Internet are not disturbed, various key performance indicators (Key Performance Indicators) need to be closely monitored. Indicator, KPI), such as CPU usage, network throughput, page views, number of online users, etc., to prevent negative impacts such as service paralysis caused by untimely detected abnormalities. Such KPIs usually have a certain seasonality, and the samples we identify as abnormal are usually points or segments that do not match the normal period, have large differences, and do not satisfy the seasonalit...

AI Technical Summary

Problems solved by technology

However, the above method only has an outlier output for the abnormality of the observed sample, and does not learn and train a distribution of the sample itself. Secondly, the existing time seri...

Images