Best Match Algorithm

Operator CoPilot recommendations are now available even when recipes do not exist for current conditions. A new best match algorithm determines the closest recipe to current conditions and notifies you that the best match algorithm was used in addition to surfacing the conditions that were changed to retrieve a recipe.

In this example, there are two conditions other than the Product: Humidity and Supplier. Humidity is a continuous condition and Supplier is a categorical condition. The Cookbook was created with the following recipes:

Unfortunately, the condition of 80% - 100% Humidity with Supplier A has never been seen before, and therefore there is no exact match recipe recommendation. In order to provide a recommendation, the best match algorithm assigns penalties to each of the available recipes to determine the closest match. Continuous condition penalties are normalized on a scale of 0-1 based on the number of possible changes. In the case of Humidity, there are 3 ranges with 2 possible changes for any given condition:

80% - 100% -> 50% - 80% results in a penalty of .5 (1 / 2 changes).

80% - 100% -> 0% - 50% results in a penalty of 1 (2 / 2 changes).

Changes in categorical conditions each result in a penalty of 1. In the case of Supplier, a change from one Supplier to another results in a penalty of 1.

A -> B results in a penalty of 1.

Looking at the possible best matches for 80% - 100% Humidity with Supplier A, it is apparent that the best match is the 50% - 80% Humidity with the Supplier A recipe, which only receives a penalty of .5.

You would be notified that the best match algorithm is in use and that the Humidity condition was changed to retrieve a recipe.