Contextual Filtering

This article covers the process to select a sub-set of PCL's to run against the ML forecasting process. This is helpful if your ML subscription has a limited forecast volume versus the total entities available to forecast.

Contextual Filtering

Contextual filtering provides the option to select specific PCLs based on your hierarchy settings from top to lower level.

The first step is to identify the Items, Customers, Locations that you would like to forecast with Machine Learning. There are two options to select these entities.

Option 1

The first option is to update the planninghierarchy.txt file and add additional ML specific levels and nodes against  items, customers, or locations that you would like to forecast. To learn how to manage and update hierarchies, please visit the article Creating a Planning Hierarchy in Advanced Business Planning.

To determine which entities to include for forecasting, you may use the forecast statistics report.

For example, in the hierarchy below we have added two hierarchical levels:

  • ML Forecast_Customer
  • ML Forecast_Item

These hierarchical levels have nodes assigned to establish if an entity should be or should not be forecasted.

If you would like to select a combination (Customer + Item, Item + Location, etc.) you could include an additional levels such as “ML Forecast_location" associated with the Location base level.

Add Newly Created Level to Forecast Hierarchy

Once ready, you may add the newly created group below the lowest level of your DC Statistical Forecasting hierarchy.

For example, if the standard forecasting process is middle out or top down with the low level being the Product Category, add the ML Forecast specific levels below the Product Category level.

This way, the ML Forecast will not impact the DC Statistical Forecasting process.

Selecting PCL Nodes to ML Forecast

In this example, we have added the level ML Forecast_Customer and ML Forecast_Item.

Clicking on the level ML Forecast_Customer exposes the nodes with that level. From there, select the "Forecast" node within the customer level.

Then within that node, you will see the ML Forecast_Item level and its nodes. Click on the Forecast node and then click on the icon. Doing so will change the color of the icon to black. This selects all the customer and item PCL's that will be forecasted. In this example, it would be all the PCLs related to Customers CUST01, CUST05, and CUST07and items 71, 120115, and 84045 within those forecast customers.

Option 2

Manually select the group that you would like to forecast. Based on your hierarchy you could select from top level or expand selection and choose specific lower level items.

In the example above we’ve selected:

  • All PCLs under “Greensburg” customer
  • All PCLs under “Salamanca” customer, “Widgets” Product category and “Atlanta DC” Location
  • All PCLs under “Georgetown #1” customer and “Widgets” Product category.


Add your comment

This site is protected by reCAPTCHA and the Google Privacy Policy and Terms of Service apply.