This tutorial simulates a Personalizer loop system which suggests which type of coffee a customer should order. The users and their preferences are stored in a user dataset. Information about the coffee is also available and stored in a coffee dataset.
Run the system for 10,000 requests and then create graph showing how fast and accurately the system learned.
Run an offline counterfactual evaluation to select an optimized learning policy, and apply that policy.
Run the system again, but for 2,000 requests and again create the graph showing the accuracy of the system.
- Azure notebooks account
- Personalizer resource
All the instructions are in the notebook. Here is an abbreviated explanation.
-
Create a new Azure notebook project.
-
Upload the files in this directory to the Azure notebook project.
-
Open the Personalizer.ipynb file and change the following values:
- The value for
<your-resource-name>in thepersonalization_base_urlto the value for your Personalizer resource - The value for
<your-resource-key>variable to one of the Personalizer resource keys.
- The value for
-
Run each cell from top to bottom. Wait until each cell is complete before running the next cell.