Managing local machine learning providers

Connect to the machine learning providers where your deployed models are stored.

Before you begin

You need the credentials of the machine learning provider you want to connect to. If you do not know where to find these credentials, contact your admin.

About this task

You can have global or local machine learning providers for your decision automations:
Table 1. Machine learning providers
Machine learning provider Description
Global machine learning providers

A global machine learning provider is accessible from all of your decision automations that are defined in its instance of Automation Decision Services and all users.

Your administrator can configure global machine learning providers on the Automation Decision Services administration page. For more information, see Configuring global machine learning providers.
Local machine learning providers

A local machine learning provider is local to your decision automation, and accessible only from your specified decision automation.

You can configure local machine learning providers on the Settings page for your decision automation in Decision Designer, as described in the procedure below.

Three remote machine learning providers are supported in Automation Decision Services: IBM Watson® Machine Learning, Amazon SageMaker, and IBM® Open Prediction Service.

IBM Open Prediction Service is an extension framework that allows you to connect to machine learning providers that are not natively supported by Automation Decision Services. This includes custom machine learning services and third-party machine learning tools, such as Microsoft Azure Machine Learning. For more information about Open Prediction Service and how to install it, see the Open Prediction Service Hub repository.

An embedded machine learning provider is also supported. This provider allows you to import any type of Predictive Model Markup Language (PMML) file and run it directly in your automation.

Procedure

  1. Click the Settings settings icon icon in the menu bar of your decision automation.
  2. Open the Machine learning providers tab and click New.
  3. Select the provider type in the drop-down list.
  4. Enter a name for the provider and optionally add a description.
  5. Complete the required fields depending on the provider type:
    Provider Credentials
    Watson Machine Learning

    Enter the following service credentials to authenticate with your Watson Machine Learning service instance:

    • API key
    • Space ID
    • Authentication URL
    • URL

    This information can be found on Watson Studio.

    The Authentication URL is populated with a default value automatically.
    Amazon SageMaker Enter the following service credentials to authenticate with your SageMaker instance:
    • Region
    • Access key ID
    • Access key value
    Open Prediction Service

    Enter the URL of your Open Prediction Service instance.
    Embedded Machine Learning

    No credentials needed.
  6. Click Test connection to verify the connection.
  7. When it is successfully connected, click Save.

What to do next

The machine learning models are now ready to be used in your decision services. You can now import the deployment of your choice to generate a predictive model template that contains all the elements to invoke a machine learning model.
Note: The IT user needs to configure machine learning providers for runtime before executing a decision service. For more information, see Configuring the decision runtime.