You are here: Basic Steps > 3. Rule Discovery > Discover Rules Task

Discover Rules Task

The Discover Rules task analyses your data to find candidate business rules for you to examine and consider adding to your rule base. A discovery specification provides parameters to control this task (see Discovery Specifications).

In order to discover rules, the Discover Rules task must be included in a session, managed within the Sessions workspace (see Sessions).

Adding a Discover Rules task to a session

ClosedAdd the Discover Rules Task to a session

  1. Navigate to the Sessions workspace.

  2. From the Navigation Pane, select a session to edit.

  3. Open the Tasks tab.

  4. Click New and select Discover Rules.

  5. In the pop-up menu that appears, select the discovery specification to be used.

    Click OK.

  6. Click  to expand the Discover Rules task.

    Specify which classes to be used, or select All Classes.

    Note: The list of classes displayed includes all data stores opened in tasks before this one in the sequence of this session.

  7. Optionally, change the value of Maximum number of objects to sample.

    Note: A larger number of objects to sample may give better results, but it will take longer to run and consume more memory in the application server.

  8. Click Save.

Generating Results

When you complete a Discover Rules task, you can view the results in HTML and XML.

The XML results are intended primarily for processing by other computer programs. The HTML results are presented in an interface that allows you to browse the results and select rules to be included in the main rule base.

To view the results, click View Results on the Sessions page after you run the Discover Rules task.

Rule Discovery HTML results

To promote a rule to the main rule base, click next to the rule and enter a name. Click OK.

Click Description, support, or lift to sort the results based on the values in that column. Lift and support are measures of the confidence the discovery algorithm has in the suggested rule. Higher numbers suggest greater confidence.