Amazon Personalize Create Solution Version

This node creates a new solution version of either a newly created solution or an existing one with Amazon Personalize. It outputs metrics of the created solution version whereby higher values are better for each of the metrics. To create a solution, you can either use one of the predefined recipes or use AutoML. There are three types of recipes. User personalization recipes build models that give item recommendations for specific users, personalized ranking recipes build models that can be used to do a personalized ranking of a list of items for a specific user, and related items recipes build models that can be used to find items that are similar to a specific item. If you are using AutoML, Amazon automatically finds the best user personalization recipe for you by trying several recipes out.
You need to have at least one dataset group that contains interactions data to be able to create a solution. Use the Amazon Personalize Upload Interactions node to upload such data.
For deployment of a solution version, create a campaign with the Amazon Personalize Create Campaign node.
More information is available on the website or in the documentation.

Note: If the node execution is interrupted during the creation, e.g. by the user or due to lost of internet connection, the creation will still be completed on AWS, although the node does not execute successfully.

Options

Create a new solution and a version of it

If selected, you can create a new version with the following settings. There will also be one version of this solution created.

Dataset group

Specify the dataset group to create a solution with. See also here.

Solution name

Specify the name of the solution. See also here.

Perform hyperparameter optimization

If enabled, Amazon will automatically try to find the best hyperparameters. This usually results in a better model but will also take more time. See also here.

Use a predefined recipe

If selected, you can select one of the predefined recipes to use. User personalization recipes build models that give item recommendations for specific users, personalized ranking recipes build models that can be used to do a personalized ranking of a list of items for a specific user, and related items recipes build models that can be used to find items that are similar to a specific item. See also here.

Use AutoML

If selected, Amazon automatically runs the aws-hrnn and aws-hrnn-metadata recipes and finds the one that performs the best. In the output you will find information about the best recipe. See also here.

Create a new solution of an existing solution

If selected, an existing solution can be selected. For this solution, a new version will be created.

Provide the solution version ARN as flow variable

If enabled, the ARN of the created solution version will be pushed out as flow variable. This can be convenient if you, e.g., want to use the solution version to create a new campaign with afterwards.

Views

This node has no views

Workflows

• 03_Amazon_Personalize_Movie_Recommendation_ExampleKNIME Hub
• Data CleansingKNIME Hub

Links

• No links available

Developers