Predict Assignment Group for incoming incidents

Use this guided template to auto-assign incoming incidents to the correct assignment group to reduce your incident resolution times.

Before you begin

Role required: piwb_manager or piwb_viewer

About this task

This template walks you through customizing a use case model to predict the correct assignment group for your incoming incidents. When the use case template shows the label Pretrained, you can bypass several implementation steps and are automatically taken to evaluate and tune your models when you click Start. Otherwise, you will begin by creating a machine learning model.

Procedure

1. Navigate to All > Predictive Intelligence Workbench > Use Cases > Create New from Template.
2. Select the Predict Assignment Group for incoming incidents guided template.

   

3. In the Use case - Predict Assignment Group for incoming incidents pop-up form, provide a unique name for your use case in the Use case name field.

   

   Note:

   The use case will contain multiple models that you create. You can use the same name as the template, if desired.
4. Although not mandatory, it's beneficial to provide unique details about the use case in the Short description field.
   Note:

   Typically the short description provides details about what the goal of the use case is or your purposes for creating the use case.
5. Provide a name for the model in the Model name field.
   Note:

   Models are trained based on default parameters. The name of the model should reflect its purpose, for example, "Model for past years data".
6. Review the base system parameters set to help train the model in the Input fields and Number of records fields.
7. Modify this data by clicking Advanced Setup.
   Note:

   By clicking Advanced Setup you can change the processing language, review and modify filters used to train the model. You can customize the filters to best represent your business data or add new criteria by clicking New Criteria. Review and modify input fields used to generate predictions. Customize the fields to best represent your business data by moving fields between the Available and Selected lists. Save any changes you make.
8. Click Start.
   A pop-up message lets you know your model is created and that training is underway.

   Note:

   To start implementation, you must provide at least a use case name and model name.

   

9. Click Done.
   The use case page displays the name and description of the use case you created. On this page, you can see all the guided implementation phases you will work through to create and implement your use case model.

   

10. Click View Progress below the header on the use case setup page to monitor a current training process.

    

    Note:

    When training is complete, you will receive an email notification letting you know your use case model training was successful. For more information about use case model notifications, refer to ITSM Predictive Intelligence Workbench notifications.
11. Click the model name under Trained Models on the use case setup page to view associated data about the trained model.
    The Model details page opens.
12. In the Retraining Schedule field of the Models detail page you can change the definition, if desired.
    The default value is Every 30 days, but you can retrain as often as once to every 180 days
13. If you made schedule retraining changes to the use case model, click either Update or Update and retrain, to retrain again.
14. To tune a model, click Tune Values on the use case setup page under Trained Models.
    The tuning page for the use case opens displaying the associated model name with the following data: Distribution (%), Net Automation, Precision (%), Coverage (%), and any comments. View the evaluation data used for the use case and associated model.

    Note:

    You can only tune values when the model is trained.
15. If desired, tune and refine the model by defining the right combination of coverage and precision to use.
    Double-click in the Precision (%) and Coverage (%) fields to change values. Changes are automatically saved.
16. If desired, click Create Another to create another model associated with your use case in the Create and train models section of the setup page.
    Repeat the process to create and train the model.
17. In the Test your models section of the use case page, click Start to begin the testing process.
    The Testing your models page opens.

    

18. Under Select models to test, decide if there are other use case models available that you want to test.
    If so, move them to the Selected list.
19. Under Define testing parameters, decide if you want to test one use case model.

    If so, select the Single test test type. Single test is the default.

    Note:

    Select Batch test when you want to test more than one use case model.
20. Determine the number of top results you want to display.
21. Under Input fields, provide a short description of your use case model test.
22. Click Run Test.
23. View the test results data for the use case model under View test results.

    

    

24. If you are ready to integrate your use case model into your business processes, return to the use case set up page and click Start in the Integrate a model section.
    The Select a model to integrate page opens.
25. In the Select a Model drop-down menu, select the model you want to integrate.
    The Select a model to integrate page opens.

    

    Note:

    You can change the retraining schedule via this page if you desire. Use the Retraining Schedule drop down selection field to change from the default 30 day value.
26. Click Integrate.
27. Click Integrate again when the pop-up asks you if you are sure you want to perform this action.
    You have integrated a use case model into your business process.

    Note:

    For details regarding trained use case integration implementation, refer to Predictive Intelligence Workbench integration and customization.