Custom Machine Learning Models

The Custom Machine Learning Models integration feature gives you the ability to deploy models and use input data from OpenPages® fields to generate live insights and suggestions in OpenPages views. This feature also gives you the ability to customize how the insights are displayed in OpenPages. Starting from version 9.1.3 and later, you can also integrate with external machine learning models using extensions.
Note: You must enable the All Permissions > SOX > User Interfaces > Show AI Insights permission for users to view insights.

To configure the model in OpenPages, you need the Custom Machine Learning Models application permission. If you don't have this permission, the Open Administration menu Administration menu > Integrations > Custom Machine Learning Models menu item is not visible.

Models that are built that use Watson Studio AutoAI or developed by a data science team can be deployed to an AI service, and integrated into OpenPages.

You can choose from the list of available AI services when you configure your model in OpenPages. The list includes services such as IBM Watson® Machine Learning on IBM Cloud®, IBM Watson Machine Learning on Cloud Pak for Data, IBM® Natural Language Understanding on Cloud, or external AI services you added through extensions.

Note: Before you use an AI service, ensure that your company selected a plan for that service that supports the volume of usage from within OpenPages.
The following models are examples of use cases that OpenPages supports:
  • Models that display the insights that are found by running the model, such as Cognitive Controls and PII models.

    Cognitive Controls models analyze text and return values for Who, What, When, Where, and Why.

    PII models detect object text that contains Personally Identifiable Information (PII).

  • Models that suggest values for fields. Fields can be set automatically or the user can set them manually.
  • Models that suggest tags to categorize resources in OpenPages. Tags can be set automatically or the user can set them manually. Tagging allows users to find different resources that are related.
The following are prerequisites for configuring Custom Machine Learning Models in OpenPages:
  • Models that are deployed on an AI service, such as IBM Watson Machine Learning, must be trained and tested by a data scientist. Testing ensures that the models have sufficient accuracies before they are deployed on the service.
  • You must be familiar with the fields used as inputs to the model, the data types of the fields, and the order in which they are defined.
  • You must be familiar with the model output format on the service.
  • You must understand which components of the model's output you want to extract.
  • You must understand JSONata (https://try.jsonata.org) syntax to extract relevant pieces of information.

When you are configuring a new model, you can save it at any time even if the configuration is incomplete, or you can cancel to exit without saving. When you have entered all of the required configuration data for a new configuration, and click Create, the configuration is validated. When the configuration successfully passes validation, it is marked as complete and can be used with views in OpenPages.

For an overview of how to use the Custom Machine Learning Models integration feature, see the following video.