Machine Learning Daily Explorer

The Machine Learning Daily Explorer (MLDE) allows you to see how specific features impact a predicted forecast volume value for a selected location, volume driver, and date combination (values are calculated for a specific driver on a specific day).

Important: Before you can view content in the MLDE, the forecast must first be generated with the appropriate explanation values (in the Parameter String field of the batch task configuration) when the volume engine is run as a scheduled batch task. For more information about batch tasks, see the Configure batch tasks online help topic.

For more information about Machine Learning in Forecasting, see the Machine Learning online help topic.

  1. To access the Machine Learning Daily Explorer, open the Forecast Planner.
  2. In the Forecast Planner (accessed from either the Volume or Forecast Home steps) make the appropriate selections to view the correct time frame, locations and forecast factors. Data presented in the Machine Learning Daily Explorer is based on these selections.
  3. Click Tools > Machine Learning Daily Explorer. The Machine Learning Daily Explorer page opens.
  4. From Location , select the appropriate location. This selection must have been used to train a machine learning model and determines the options available in the Volume Driver drop-down menu.
  5. Select the appropriate volume driver.
  6. Select the appropriate date. Dates shown depend upon the forecast week selected in the Forecast Planner.
  7. Click Apply. The Apply button is disabled until you select new parameters.
    • To view another feature contribution graph, select new parameters and click Apply.
    • To exit the Machine Learning Configuration page, click Done or X.

Any forecast that you want to view in the MLDE must have been generated using machine learning. In addition, 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 Daily Explorer, you cannot access the Machine Learning Daily Explorer.

After clicking Apply in the Machine Learning Daily Explorer configuration page, view the Contribution by Each Feature bar graph. The graph depicts the impact of each feature on the prediction, using the values that the machine learning process calculated for a specific driver on a specific day.

What's in the Contribution by Each Feature View?

The Contribution by Each Feature view includes the following:

  • For your reference, the selected machine learning configuration appears just under the page title.

  • The X-axis displays each feature name.

  • The Y-axis displays the range of predicted values with the predicted value for the selected configuration appearing in red next to the red dashed line.

  • The base value and predication value appear just below the chart header .

  • Each bar represents how much is being added or subtracted from the base value. At the end, the bar for the final feature in the chart should land on the dashed prediction line.

    Hover over each bar graph entry to view its data.

    • The pink entries represent values that subtract from the base value.
      (Prediction = base value - sum).

    • The blue entries on the graph represent values that add to the base value.
      (Prediction = base value + sum).

  • The red dashed line represents the prediction value.

  • If the selected MLE configuration does not have the appropriate data to generate a graph, the system generates an error.