Manage a predictive model

Use the following procedure to create a predictive model and define when to evaluate the score during a process. You can access the predictive score for a job within the process itself using the APIs.

Create a predictive model

Define data and the corresponding weighting and scoring system (score rule) for a predictive model.

  1. On the process properties panel, click Predictive model.
  2. Click for Weightings.
    The New weight dialog box is displayed.
  3. Configure the following weighting properties.

    Variable

    Select a variable to define a predictor, such as Age. If the required variable is not available, you can create a new variable directly from the process properties panel. See Create a process variable while configuring process properties.

    Operator

    Select an operator, such as Equals.

    Alternatively, if you need to use a range of values, such as age greater than 25 and less than equal to 30, select Range as the operator.

    Value

    Enter a value for the variable, such as Male for Gender and 25 for Age.

    Weight

    Enter a fixed value for the weight of the variable.

  4. Click Add.
  5. Add other variables, values, and weights as needed.
    Weightings cannot be added for XML and JSON expression variables.
  6. Define the Score rule to use for scoring.
    1. Right-click on the rule editor and select predefined variables to create the rule, such as Gender+(Age*2).
    2. Click Validate.

    • If you document a process that has a predictive model, the weighting and score calculation for the process are included in the documentation.
    • If you export or import a process that has a predictive model, the weighting and score calculation for the process are included in the exported or imported process.

Evaluate the predictive model score within a process

Indicate when to evaluate the score in a process by defining points of score evaluation. For example, in an Insurance Sales process, evaluate the score after all the information for the sale of an insurance policy is collected.

  1. Select the activity for which to evaluate the predictive model score.
  2. In the properties panel of the activity, click the Advanced tab.
  3. To evaluate the predictive model score when the activity becomes available or completes, select Evaluate score.
  4. Release the map.

    When importing and exporting a process, the score for an activity is also imported and exported. The predictive model scores are also included in the documentation for the node.

Access the predictive score for a job

You can access the predictive score for a job within the process itself using the APIs. For example, use the score within a decision to determine the path of execution for the process.

  1. Create a process variable of type System with a value, JOB_SCORE. This variable holds the score value on score evaluation.
  2. Use the following APIs:
    • GetJobProperties API on the JobService to view the job score at runtime.
    • ActivityService.GetWorkqueue API to filter activities on a work queue based on a particular score.
    • Evaluate Score API on the JobService to reevaluate the score.
    For more information on these APIs, refer to the API Online Help on the Product Documentation page.