Anomaly Detection Function

Description

The Anomaly Detection Identifies outlier events and the start of new trends in the data.

Configuration

To add the Anomaly Detection Function to a stream, follow the steps below:

1. Ensure you have a use case open in the stream designer, this can be a new use case or an existing use case
2. From the toolbox on the left expand the Function option and scroll down until you can see “Anomaly Detection”
   • You can also use the search in the header to find the stream object quickly
3. Click and drag the Aggregate Transformation from the toolbox onto the canvas
4. Rename the agent by clicking into the text button on the action bar
5. Save the Stream by clicking the save button in the action bar
6. Hover over the icon for the stream object until it turns orange and then double click to open the configuration page
   • You can optionally use the configure option on the action bar once you have selected the specific stream object and then clicking this option
7. Configuration options
   • The drop-down allows you to associate this stream object with a specific collection. The default option is derived from the default specified against the specific use case and is generated left as the same option. If you do need to change it to another collection make the change by selecting the new collection in the drop-down.
8. Enter Training details
   • Select Learning Algorithm from drop-down list
   • Attach Training File by clicking Select File or Drop file
   • Tick “Has Header?”
   • Select Separator Character from drop-down list
   • Select Input Field from drop-down list
   • Select Input Map from drop-down list
9. Enter Advanced Options
   • Enter Sensitivity, which accepts values between 0 and 100
   • Enter Change History Length
10. Enter Model options
    • Enter Deterministic Seed
11. Click Apply on the action bar, and then save the stream using the save button.

Limitations

None at this time