Real-Time Intelligence tutorial part 7: Detect anomalies on an Eventhouse table

In this part of the tutorial, you build a real-time data workflow that detects anomalies in streaming data. You use an Eventhouse table to analyze time series data and identify unusual patterns.

Create an anomaly detector

Open the Eventhouse table created in the previous tutorial.

1. From the left navigation pane, select Real-Time.

   

2. Under Streaming data, select the TransformedData table.

   

3. On the table details page, select Detect anomalies from the toolbar.

   

Create a detector to analyze the data for anomalies.

In the New Anomaly detector pane:

1. Enter a name for the detector.
2. Select your Fabric workspace.
3. Select Create.

Configure anomaly detection

Configure the attributes used to detect anomalies.

1. In the Edit configuration section, set the following values:

   Field	Value
   Value to watch	No_Bikes
   Group by	Street
   Timestamp	Timestamp

   

2. Select Save.

Choose an anomaly detection model

1. In the Find models section, select Analyze my data to find the best anomaly detection model for your data.

   

2. Review the recommended models and select the one that best fits your needs. For this tutorial, select the recommended Local Pattern Detector model.

   

3. Select Save.

Review anomaly results

After the analysis completes, review the detected anomalies.

1. View the anomaly results in the Detector results pane.

2. Inspect the chart and tabular output to identify unusual patterns.

   

Next step