Custom dataset groups - Amazon Personalize

Custom dataset groups

With Custom dataset groups, you build custom resources for configurable use cases. You train, and deploy configurable solutions and solution versions (a trained Amazon Personalize recommendation model) based on your business needs. The Custom dataset groups workflow is as follows:

  1. Format your historical input data based on custom schema requirements in Custom datasets and schemas and upload the data into an Amazon S3 bucket.

  2. Import your data into an Amazon Personalize dataset and record real-time event data from user interactions with items.

  3. Choose a recipe (algorithm) to use on the data and when you have collected enough data, train a solution version with the recipe.

    • At minimum 1000 interactions records from users interacting with items in your catalog. These interactions can be from bulk imports, or streamed events, or both.

    • At minimum 25 unique user IDs with at least 2 interactions for each.

    Different recipes have additional data requirements. For more information see Step 1: Choosing a recipe.

  4. Get recommendations or user segments with the following methods:

    • Deploy the solution version with a campaign and get recommendations.

    • Create a batch inference job to get batch recommendations.

    • Create a user segment job to get user segments.