The built-in machine learning–based operator ANOMALYDETECTION is designed to help customers of Azure Stream Analytics who monitor data from applications or devices in real time, and who need help to easily detect events or observations that do not conform to an expected pattern.
The complexity of machine learning has been simplified to a single SQL function call to a machine learning model. The underlying general-purpose machine learning model is abstracted and continuously learns over time to match your input streams.
This functionality is targeted toward numerical time series data. It can help you detect anomalies, positive trends, and negative trends.
You can find more information in the documentation about anomaly detection.