Edit the Number of Events Required to Build an ART Model

If there are less than 100 events available for your Automatic Relevance Tuning (ART) model, it is empty and does not provide recommendations, as there is not enough data to learn from (see Managing Coveo Machine Learning Automatic Relevance Tuning Models in a Query Pipeline and Evaluating if Your Search Interface Produces Enough Data for ART). With 100 events or more, your model starts learning and improving. However, it is possible to change this threshold if needed.

  • You create an ART model when very few events have been gathered. You want your to start providing recommendations only when there is sufficient data to learn from and make good recommendations for the most frequent queries. You therefore set the threshold to 3,000.

  • You create an ART model for testing purposes. Since you are the only user generating events, you want your model to provide recommendations immediately, i.e., with just a few events. You therefore set the threshold to 0.

To edit the number of events required to build an ART model:

  1. If not already done, log in to your Coveo Cloud organization with an account member of a group with the appropriate privileges (see Required Privileges).

  2. Access the Query Pipelines page (see Access the Query Pipelines Page).

  3. In the Query Pipelines page, double-click the pipeline in which your ART model applies.

  4. In the Machine Learning tab, click the desired model, and then, in the Action bar, click Edit Code.

  5. In the Edit a Model with Code panel, add the following to the existing code, and then replace 100 by the desired threshold.

     customModelParameters: [“--conf”, “coveo.drill.minEventsForModelBuilding=100"]