When the advanced ML or analytic algorithms are used to apply outlier, clustering, regression, forecasting, or fill blanks calculations to a query, an explanation is automatically generated and can be shown in your visuals as Notes (which you can see in Discover or in runtime presentations) and in Chatbot responses when calculations are added.
These explanations are generated automatically to help you to understand or communicate the calculations used in your queries. They typically contain the calculation name, the query attributes, and the method (algorithm) used to calculate the results.
Note: The explanatory notes are auto-updated whenever the underlying mechanism is changed.
Showing Auto-Explanations
You can see the automatically generated notes:
- In Discover, by selecting Show Notes in the Design ribbon (orange arrow). These auto-generated notes are appended to any custom user-generated notes. For more information, see Showing Notes in Discover
- In runtime presentations, where a Notes icon is added to the embedded visual. This functionality is downstream of the discovery; that is, the Show Notes option must be enabled in Discover for the icon to appear in your presentation and its content (opened on hover) will match the content added in Discover. For more information, see Explanations at runtime
- In the Chatbot, where the explanation is presented in the response when a calculation is added. For more information, see Explain the results.
Example: Showing an explanation for Outliers
Example: Creating a Forecast using the Chatbot
Another example is shown below for forecasting. In this case, the Chatbot was used to generate the forecast and the Show Notes option has been enabled for the visual.
In this example, the Notes panel is shown (orange arrow below) and has been repositioned to the side of the visual using drag-and-drop (blue arrow). You can see the explanation in the Notes panel (purple highlight) and the Chatbot response (blue highlight).
Note: You do not need to be showing notes for the Chatbot to tell you how the forecast was generated in its response.
