Adding or Updating Models

To add or update a Ticket Classification model:

From the Ticket Classification Models view, do one of the following:
- To add a new model, click Add.
- To edit an existing model, search for or navigate to the model, and then click the model name.
Provide a name for the model. Note that you must create separate models for each field that Work Center will predict on a transition form; therefore, consider including the name of the affected field when adding a new model.
Specify the Application. This determines which project, fields, and transitions apply to the model.
Select a Project. Search for a project or select it from the Project Hierarchy. Select Include Sub-Projects to broaden the scope of available fields and transitions.
Select the Input Fields that the system will use to predict field values. Note that only Text fields in the selected projects are displayed.
Select the Output Field that is affected. Selection, User, and Single Relational fields are available as output fields. Note that only one field is available per model. This means that you must create a separate model for each field that can be predicted on a transition form.
In the Training Dataset Filter, optionally click Add to define search parameters to filter the items that are used to train the model. This trains the model with items that are pertinent to where this model will be applied.
For example, if you want the model to predict values only based on QA tickets in the ACME project, create a search filter like:
```
Item Type contains any QA
and
Project contains any ACME
```
To improve prediction accuracy, train the model using a broad data set, but try to limit unrelated items that may skew the results. For example, consider training model to exclude items in which the Functional Area field is empty. In this case, you can add the filter:
```
Functional Area does not contain any (None)
```
Select the transitions that apply. This determines which transitions will display the Predict Field Values button and icon () on the transition form. You might use the same model on multiple transitions, or you might use distinct models for each transition as necessary.
Tip: You can use the Run Transition action in a notification to run a transition and predict a field value based on text in the item.
1. For example, create a model with the following mappings:
  - Input – Title, Description
  - Output – Developer
2. Select a transition that will be executed automatically like "Assign to Developer".
3. Train the model.
4. In the "Any Item is Submitted" notification, set the Action to Run Transition, and then select the "Assign to Developer" transition.
5. Click the Mappings button. Select (Auto) for the Developer field.
After an item is submitted, the Assign to Developer transition is executed automatically and the Developer field value is set based on the text the submitter entered in the Title and Description fields.
Save your changes.

When you are finished adding or updating the model, you can click Train to update the system immediately. However, due to potentially high CPU-usage that can occur when training a model, you might consider training the model during off-peak hours. For details, refer to Training a Model.