What is Built-in Machine Learning-based Anomaly Detection?
The customers of Azure Stream Analytics can easily detect events or observations with the help of built-in operator ‘ANOMALYDETECTION’. The users can detect data in real time and conform the data to the expected patterns. The single machine learning models are used for the streaming of data and it requires a complete understanding of various scenarios and models. The complexity of the system has removed with a single SQL function which has removed the need for complex data pipeline engineering. With the help of general purpose machine learning model, the user can match different input streams and the functions are targeted towards the numerical time data series. Anomaly Detection documentation provides the complete details of positive and negative anomaly trends. Here are some examples of anomaly data series:
Here is how you can use multiple ways to use anomalies:
Egress to Azure Functions
Azure Functions is a serverless compute service which is highly useful for the customers to run the code triggered by the events occurring in Azure or third party services. The output of the Azure functions is the output of the Azure Stream Analytics and one can use these services in writing the output to service bus queue. Output Adaptor is introduced to integrate Azure functions and Azure Stream Analytics. It connects both Azure functions and Azure Stream Analytics as well as run a script to trigger the downstream workflow. This is useful for the developers to send a notification when the system exceeds the pre-defined values.