Machine Learning Model Explorer

The Machine Learning Model Explorer shows a graphical break down of how different Machine Learning configurations generate a forecast, specifically the way in which specific features impact the final calculation.

The Machine Learning Model Explorer provides two different views:

  • Feature Importance - Displays a bar graph showing how each model feature has contributed to the model. The graph displays each feature in descending order based on a feature's impact to the model.
  • Feature Dependence - Displays a scatter plot that shows the dependency between a specific feature and its value. After selecting a feature from the drop-down menu, the scatter plot displays the trend and shows whether the feature has a positive, neutral or negative impact on the forecast.
For more information about Machine Learning in Forecasting, see the Machine Learning online help topic .
  1. To access the Machine Learning Explorer, open the Forecast Planner.
  2. In the Forecast Planner, make the appropriate selections to view the correct time frame, locations and forecast factors. Data presented in the Machine Learning Explorer is based on these selections.
  3. Click Tools > Machine Learning Model Explorer. The Machine Learning Model Explorer page opens.
  4. From Machine Learning Configuration Name, select the appropriate name of a machine learning configuration. This selection must have been used to train a machine learning model and determines the options available in the next menus.
  5. Make the appropriate selection. Depending on the pooling strategy assigned in the selected machine learning configuration, this menu will display the location for the model. This is based on the locations that are selected in the Forecast Planner.
  6. If the pooling strategy in the selected machine learning configuration is by:
    • By Driver, the menu is disabled and the name of the root location corresponding to the locations selected in Forecast Planner is pre-selected.
    • Driver and Region, the menu label changes to Region and displays a list a list of the region names corresponding to the locations selected in Forecast Planner is available for a selection.
    • By Driver and Store, the menu label changes to Store and displays a list of the store names corresponding to the locations selected in Forecast Planner.
    • Driver and Department, the menu label changes to Generic Department and displays a list of the generic departments corresponding to the locations selected in Forecast Planner.
  7. Select a volume driver. Different models are created for each volume driver.

Note: If no Machine Learning configurations are present on the system or have not been used in machine learning training when you try to access the Machine Learning Explorer, you cannot access the Machine Learning Explorer.

After clicking Apply in the Machine Learning Global Explorer configuration page, select the appropriate tab:

  • Feature Importance - Displays a bar graph showing how each model feature has contributed to the model.
  • Each feature’s weight or amount of contribution to the model is depicted as a bar and appears in descending order. The graph displays the importance of each feature in the model creation based on the value calculated by the Machine Learning Explorer on all the data or a sample of data for each volume driver separately.
  • Feature Dependence - Displays a scatter plot that shows the dependency between a specific feature’s value and its corresponding explorer value. The plot depicts whether a feature’s impact on the model is positive (trending upward), negative (trending downward) or neutral (straight line).
  • To view a specific feature’s impact, select the appropriate feature from the Feature drop-down menu.