Skip to main contentIBM Garage Digital Business Automation Reference Architecture

Solution configuration

This chapter documents the environment pre-requisites to deploy the artifacts supporting the integration of WML with ODM.

Environment

In order to run the use case, you will need to have an OpenShift cluster with the following components installed:

  • IBM Cloud Pak for Data 3.0.1
  • IBM Cloud Pak for Automation version 20.0.2
    • Operational Decision Manager (ODM)
    • Business Automation Insights (BAI)

Configuration

Deployment

WML scoring service deployment

In order to run a scoring service from ODM, you need to create and deploy this scoring service to WML. The easiest way to achieve this is to create a service using the AutoAI capability of CP4D, following these steps:

  • Open a CP4D project and click the Add to project button, then select the AutoAI experiment asset type.
  • Provide a name for you experiment, e.g. AutoAI GCR.
  • On the next screen, when prompted for a data source, drag and drop the training data CSV file gcr-training-data.csv from the ml-training-data folder to the Add data source pane. The data is then loaded and the available data columns are displayed.
  • In Select prediction column, select the Risk column. At this point, you should see the following: autoai experiment setting
  • Click the Run experiment button. When the generation of alternative pipelines complete, select one of them and click Save as > Model, then save the model. save pipeline
  • When prompted by a green message box, click View in project and on the model page, click the Promote to deployment space link.
  • From the main menu (hamburger on the left), you can select Analytics deployment then select the deployment space you used to deploy your newly minted scoring model. You should now see something like this: deployment space
  • Click on your model, then on the next page, click the Create deployment button. Select Online for the deployment type, provide a name for the deployment and click Create. After a while, your scoring service is now deployed as a web service. Click on the service to get the endpoint to execute the scoring. Take note of this endpoint as it is the one that you will provide to the ODM decision service so it can invoke the scoring service.

You can also perform some quick, interactive tests of the scoring service from this page.

deployed service details

ODM decision services deployment

As mentioned in the use case definition, there are several possible approaches to the design of the object model for the rules. Two decision service archives, a hierarchical and flat version, are available under the exports folder. You can load them in ODM Decision Center by following these steps:

  • Open the Decision Center business console and click on the Library tab.
  • Select the Import Decision Service button, navigate to the dc-export - risk-assessment-main.zip archive in the exports folder and click the Import button. Repeat for the same process for the dc-export - dynamic-risk-assessment-main.zip archive. The decision service projects are now loaded: loaded decision services