Interactive scoring

Interactive scoring, or “what-if” analysis, explores how a time series forecasting model behaves under different conditions. Create and compare multiple “scenarios” by modifying external feature values to assess how changes impact the model’s forecasts.

From the model’s result page, view forecasts for different scenarios on a chart and edit the external features for each scenario in a table. The table for editing scenario data is similar to an editable dataset.

To analyze different outcomes, create multiple scenarios. Each new scenario can start from the last test horizon or from any other scorable timestamp in the dataset. To build upon existing data, duplicate and then modify a scenario. Scenarios can be renamed for clarity or deleted when no longer needed.

Publishing to a dashboard

Publish the interactive scoring interface to a dashboard as part of a saved model report. The published report includes both the forecast chart and the data table for your scenarios.

Limitations

Interactive scoring is only available for models with “Known-in-advance” features.