Enabling field value recommendations

You can use the AI configuration application to enable AI-recommended values for fields. The field value recommendations use the mcc model template. This configuration does not support problem code recommendations in work orders. That AI feature requires a separate configuration.

1. In Maximo Manage, in the AI configuration application, click Add configuration.
2. In the Name field, specify a name
3. Optional: In the Decription field, specify a description.
4. In the Template field, select the mcc template.
5. In the Template version field, select the latest template version.
6. In the Object structure field, select the object structure that you used in the invocation channels and the training and inference filters.
7. Optional: In the Target object path field, select the hierarchy path for the child object.
   This value is required only when the inference is on an attribute of a child object and not the main object of the object structure. When you select child objects, during training and inferencing, the target object path maps the hierarchy path in the object structure. The hierarchy path supports only the immediate child to the root object.
8. In the Attribute field, specify the specific object structure attribute that represents the field for which you want to generate recommended values. Select the same attribute as you specified in the invocation channel request template for training.
9. Optional: In the Target description field, specify the name of the attribute that contains the description for the target attribute.
   For example, if the target attribute is for work types, specify the description field object for work types. By selecting the target description, the model can more effectively recommend values.
10. In the Training invoke channel field, select the invocation channel for training.
11. In the Training filter field, select the training filter.
12. In the Inference invoke channel field, select the invocation channel for inferencing.
13. In the Inference filter field, specify the inferencing filter.
14. Click Create.
15. In the AI configurations table, select the AI configuration that you created.
16. Set up arguments.
    Arguments control some aspects of the models output, such as the threshold for acceptable values.
    1. Click Actions > Set arguments.
    2. Specify a value for the arguments. The following table describes the arguments. You must set a value for the features argument.

       Table 1. Arguments for mcc model template

       Argument key	Description	Type	Default value
       features	Specify the description-related features that you specified in the query templates for the object structure. Specify a comma-separated list. If the target attribute description is sourced from a related object, specify only the attribute description. To specify a description that is a child attribute of multiple related objects, use the following notation: relationship_name.attribute_name*	string	No default value
       score _threshold	The model generates a score that measures how recommendable each value is. Values that have scores above the threshold are considered for recommendation. If you are setting up this type of recommendation for the first time, you might choose to set the score threshold to a lower value. A lower value increases the likelihood of any output from the model, although it does typically decrease desirable output, especially if the training filter does not contain diverse or adequate amounts of data.	integer	0.5

    3. Click Save.
17. Optional: In the Edit AI configuration dialog, specify information about your model.
    1. Click Actions > Edit.
    2. In the Additional details for AI explained section, provide any information that is specific to your organization and relevant to your end users to help them understand the model and its output.
       An AI icon is located alongside your model's output. Your users can click the AI icon and then access any information you specify in this section alongside other general model information that's provided by IBM. You can complete the AI configuration process, review the AI icon and its content in context, and then edit this section later as needed.
    3. Click Save.
18. Click Actions > Check data requirement.

    When you check the data requirements, the training data is reviewed to determine if the data contains enough detail. If the data check fails, you must add or improve the quality of data in your training filter. For more information, see Preparing required components for field value recommendations.

19. Click Actions > Activate

    Activating the AI configuration indicates that the AI configuration is prepared and the model is ready to be trained.

20. Optional: Change the frequency of the training process

    Training is controlled on a crontask schedule. The AITRAINJOB crontask initiates training for all eligible AI configurations. By default, the crontask runs every five minutes.

    For more information, see Cron tasks for training and inferencing.

    If you want training to run sooner, you can edit the cron task schedule.

    1. In Maximo Manage, in the Cron Task Setup application, open the AITRAINJOB cron task.
    2. In the Cron Task Instances table, for the WOAI instance, in the Schedule field, change the value.
    Wait a few minutes before continuing to the next step.
21. In the AI configuration application, in the AI configuration for field value recommendations, click Actions > Train model

    Training begins when the AITRAINJOB crontask runs. Training can take a few hours.

    For conceptual information about training, see Model training overview.

    You can monitor training in the Model training log table or in the Model status dialog.

    The Model training log table is in the AI configuration that you created and contains step-by-step updates for the training process, but you must refresh the page to see updates. To refresh the page, click Refresh.

    To access the Model status dialog, in the AI configuration, click Actions > Check model status.

    The model accuracy score is a measure of how the model performs on the training data. The score represents the amount of values that the model recommended that it considers reasonable to be the correct or best value. The closer to 1, the more accurate the output likely is in context of the data on which the model was trained. If the model was trained on data that was not complete or diverse but the score threshold is low, the accuracy score might be high but the output is not accurate.

    If training fails, you can complete some troubleshooting steps. For more information, see Troubleshooting Maximo AI Service and AI features.