Събитие
31.03, 23 ч. - 2.04, 23 ч.
Най-голямото събитие за обучение на Fabric, Power BI и SQL. 31 март – 2 април. Използвайте код FABINSIDER, за да спестите $400.
Регистрирайте се днесТози браузър вече не се поддържа.
Надстройте до Microsoft Edge, за да се възползвате от най-новите функции, актуализации на защитата и техническа поддръжка.
What is multivariate anomaly detection for time series? Univariate anomaly detection, which is implemented by the KQL function series_decompose_anomalies(), enables you to monitor and detect anomalies in the distribution of a single variable over time. In contrast, multivariate anomaly detection is a method of detecting anomalies in the joint distribution of multiple variables over time. This method is useful when the variables are correlated, thus the combination of their values at specific time might be anomalous, while the value of each variable by itself is normal. Multivariate anomaly detection can be used in various applications, such as monitoring the health of complex IoT systems, detecting fraud in financial transactions, and identifying unusual patterns in network traffic.
For example, consider a system that monitors the performance of a fleet of vehicles. The system collects data on various metrics, such as speed, fuel consumption, and engine temperature. By analyzing these metrics together, the system can detect anomalies that wouldn't be apparent by analyzing each metric individually. On its own, an increase in fuel consumption could be due to various acceptable reasons. However, a sudden increase in fuel consumption combined with a decrease in engine temperature could indicate a problem with the engine, even if each metric on its own is within normal range.
Multivariate anomaly detection in Fabric takes advantage of the powerful Spark and Eventhouse engines on top of a shared persistent storage layer. The initial data can be ingested into an Eventhouse, and exposed in the OneLake. The anomaly detection model can then be trained using the Spark engine, and the predictions of anomalies on new streaming data can be done in real time using the Eventhouse engine. The interconnection of these engines that can process the same data in the shared storage allows for a seamless flow of data from ingestion, via model training, to prediction of anomalies. This workflow is simple and powerful for real-time monitoring and detecting of anomalies in complex systems.
This solution relies on the following components:
Събитие
31.03, 23 ч. - 2.04, 23 ч.
Най-голямото събитие за обучение на Fabric, Power BI и SQL. 31 март – 2 април. Използвайте код FABINSIDER, за да спестите $400.
Регистрирайте се днесОбучение
Пътека за обучение
Get started with Microsoft Fabric - Training
Microsoft Fabric learn modules, Fabric learn modules, learn Fabric, Fabric training, Microsoft Fabric training, Microsoft Fabric learning paths, Get started with Microsoft Fabric
Сертифициране
Microsoft Certified: Fabric Analytics Engineer Associate - Certifications
As a Fabric analytics engineer associate, you should have subject matter expertise in designing, creating, and deploying enterprise-scale data analytics solutions.
Документация
Multivariate anomaly detection - Microsoft Fabric
Learn how to perform multivariate anomaly detection in Real-Time Intelligence.
Analyze time series - Microsoft Fabric
Use SynapseML and Azure AI services for multivariate anomaly detection.