PlanIQ - How to use item attributes to refine your forecast

edited December 2022 in Best Practices

Metadata attributes are static, non-time dependent categorical text features that describe the items in the historical time series data. Examples for attributes could be style, category, geographic location, size, item hierarchy level etc. The advanced neural network machine learning algorithms supported by PlanIQ can leverage the information captured in these attributes to produce more accurate forecasts. They do so by discovering patterns across the available time series data based on similar items.

Attributes are extremely useful in cold start / NPI (New Product Introduction) scenarios, where forecasts should be produced for new items for which there is limited historical data, or no historical data at all. In instances where those items share attributes (characteristics) with preexisting items, the neural network machine learning algorithms can use the attributes to utilize similar items’ historical data to produce forecasts for the new items. For more information,.

In order to use attributes, users must select either DeepAR+ or CNN-QR as the forecast model algorithm, and train the model based on a data collection that contains an attributes module.  Please note the following:

  1. Every item in the historical data module should be present in the attributes module.
  2. The attributes module can contain up to 9 attribute fields. PlanIQ assumes those fields to be text (string).
  3. There is no naming convention for these fields.

For more information about the attributes module, please refer to Anapedia.

Example of an Attribute ModuleExample of an Attribute Module


Got feedback on this content? Let us know in the comments below.


Contributing authors: Nitzan Paz, Christophe Keomanivong, Frankie Wolf, Timothy Brennan, Andrew Martin, and Evgenya Kontorovich.