English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Multivariate time series anomaly detection has been extensively studied underthe semi-supervised setting, where a training dataset with all normal instancesis required. However, preparing such a dataset is very laborious since eachsingle data instance should be fully guaranteed to be normal. It is, therefore,desired to explore multivariate time series anomaly detection methods based onthe dataset without any label knowledge. In this paper, we propose MTGFlow, anunsupervised anomaly detection approach for multivariate time series anomalydetection via dynamic graph and entity-aware normalizing flow, leaning only ona widely accepted hypothesis that abnormal instances exhibit sparse densitiesthan the normal. However, the complex interdependencies among entities and thediverse inherent characteristics of each entity pose significant challenges onthe density estimation, let alone to detect anomalies based on the estimatedpossibility distribution. To tackle these problems, we propose to learn themutual and dynamic relations among entities via a graph structure learningmodel, which helps to model accurate distribution of multivariate time series.Moreover, taking account of distinct characteristics of the individualentities, an entity-aware normalizing flow is developed to describe each entityinto a parameterized normal distribution, thereby producing fine-graineddensity estimation. Incorporating these two strategies, MTGFlow achievessuperior anomaly detection performance. Experiments on five public datasetswith seven baselines are conducted, MTGFlow outperforms the SOTA methods by upto 5.0 AUROC\%. Codes will be released athttps://github.com/zqhang/Detecting-Multivariate-Time-Series-Anomalies-with-Zero-Known-Label.

Detecting Multivariate Time Series Anomalies with Zero Known Label