Use case and recipe features - Amazon Personalize

Use case and recipe features

With some use case and recipes, Amazon Personalize uses the following features to generate more relevant recommendations and improve item discovery and engagement.

Real-time personalization

With some use cases and recipes, Amazon Personalize uses real-time personalization to update and adapt recommendations according to a user's evolving interest. It updates recommendations for a user as you record their interactions with items or actions present at the latest full training. You record these interactions with an event tracker and the PutEvents operation or, for interactions with actions, the PutActionInteractions operation.

For more information about recording events, see Recording events. For information about new data influences real-time recommendations, including real-time personalization, see How new data influences real-time recommendations.

The following use cases and recipes support real-time personalization:

Exploration

For some domain use cases and custom recipes, Amazon Personalize uses exploration when recommending items. With exploration, recommendations include some items or actions that would be typically less likely to be recommended for the user, such as new items or actions, items or actions with few interactions, or items or actions less relevant for the user based on their previous behavior. This improves item discovery and engagement when you have a fast-changing catalog, or when new items, such as news articles or promotions, are more relevant to users because they are fresh.

If your use case or recipe uses exploration, when you create a recommender or custom campaign, or when you create a batch inference job (custom resources), you can configure exploration with the following fields:

  • Emphasis on exploring less relevant items (exploration weight) – Configure how much to explore. Specify a decimal value between 0 to 1. The default is 0.3. The closer the value is to 1, the more exploration. With more exploration, recommendations include more items with less item interactions data or relevance based on previous behavior. At zero, no exploration occurs and recommendations are based on current data (relevance).

  • Exploration item age cutoff – Specify the maximum item age in days since the latest interaction across all items in the Item interactions dataset. This defines the scope of item exploration based on item age. Amazon Personalize determines item age based on its creation timestamp or, if creation timestamp data is missing, item interactions data. For more information how Amazon Personalize determines item age, see Creation timestamp data.

    To increase the items Amazon Personalize considers during exploration, enter a greater value. The minimum is 1 day and the default is 30 days. Recommendations might include items that are older than the item age cut off you specify. This is because these items are relevant to the user and exploration didn't identify them.

Use cases and recipes that use exploration

For more information about each use case or recipe that uses exploration, see the following:

Automatic updates

For some use cases and custom recipes, Amazon Personalize automatically updates your recommender or solution version to consider new items or actions for recommendations. There is no cost for automatic updates. For a list of use cases and recipes with automatic updates, see Domain use cases and custom recipes with automatic updates.

Automatic updates work as follows:

  • When Amazon Personalize automatically updates your solution version or recommender depends on how you get recommendations:

    • For real-time recommendations, Amazon Personalize updates the solution version or recommender every two hours.

    • For batch item recommendations, when you create a batch inference job and specify the latest fully trained solution version for your solution, Amazon Personalize automatically updates the solution version to consider new items during exploration. If you don't specify the latest solution version, no update occurs.

  • With each update, Amazon Personalize starts including new items in recommendations using Exploration. When considering a new item or action, Amazon Personalize considers any metadata for the item. However, this data will have a greater effect on recommendations only after you record interactions for the item and fully retrain.

  • Automatic updates are not a full retraining. Instead, automatic updates allow Amazon Personalize to feature your new items in recommendations before your next full retraining. A full training can be after your domain recommender's weekly automatic retraining completes. Or it can be after you create a new solution version with trainingMode set to FULL.

  • For an update to occur, you must provide new action, item, or interactions data since the last automatic update or full retraining.

  • Amazon Personalize considers new items until you import 750,000 items. This is the maximum number of items that are considered during training.

Additional guidelines and requirements for custom resources

If you use custom resources, the following are guidelines and requirements for auto updates:

  • Your solution version must be deployed in a campaign. Your campaign automatically uses the updated solution version.

  • Automatic updates are not a full retraining. You should still train a new solution version weekly with trainingMode set to FULL so the model can learn from your users' behavior.

  • If you manually create a new solution version, Amazon Personalize will not automatically update older solution versions, even if you deployed them in a campaign.

  • If every two hours is not frequent enough, you can manually create a solution version with trainingMode set to UPDATE to include those new items in recommendations. Just remember that Amazon Personalize automatically updates only your latest fully trained solution version. The manually updated solution version won't be automatically updated in the future. To resume updates, create a new solution with training mode set to FULL and deploy it in a campaign.

Domain use cases and custom recipes with automatic updates

For more information about each use case or recipe that features automatic updates, see the following: