Generating training data for the cognitive service dependent on the Helix model

Leverage AI capabilities provided by the Helix model in BMC Helix ITSM to predict service disruption, incident categorization, and support groups for faster resolution. For improving the overall incident management, we recommend that you train the models of the following cognitive service capabilities:

Service prediction
Categorization prediction
Support group prediction

Important
The creation of a new Multi-Attribute Prediction data set is no longer supported. However, the existing dataset, if already created, remains usable for predictions.

Before you begin

Make sure that you have the Cognitive Service Config permission. For information about this permission, see Permissions-to-configure-the-cognitive-service.
Make sure that your organization has sufficient closed incidents to train the model. Training the model on a high number of closed incidents may improve the prediction.

To generate training data by using Data Qualifier as an input source

Navigate to Application Administration Console > Custom Configuration > Foundation > Advanced Options > Cognitive Service Configuration.
On the Cognitive Data Setup form, select the required Cognitive Data Type, and perform the steps indicated in the following table:

Cognitive data type	Action
Service Prediction	From the Company list, select the company for which you want to enable AI-based service impact prediction. From the Application list, select Incident Management. From the Template list, select Incident Service CI Training Data. Click Add Selected Template. Select the added template and click Generate Training Data. The AI algorithm analyzes closed incidents from the selected company in BMC Helix ITSM, and considers them for model training. The time taken to generate training data depends on the number of closed incidents on which the training data is generated. If the training data generation fails, see Troubleshooting incident requests
Categorization	From the Company list, select the company for which you want to enable categorization. From the Application list, select Incident Management. From the Template list, select one of the following values: Select Incident Categorization Training Data to train the data for the Operational category and Product category fields. Select Incident Resolution Categorization Training Data to train the data for the Resolution category and Resolution product category fields. Click Add Selected Template. Select the training data record from the list of training data sets. Click Generate Training Data.
Support Group Prediction	From the Company list, select the company for which you want to enable AI-based support group prediction. From the Application list, select Incident Management. From the Template list, select Incident Support Group Prediction Training Data. Click Add Selected Template. Select the added template and click Generate Training Data. The AI algorithm analyzes closed incidents from the selected company in BMC Helix ITSM, and considers the incidents closed within the past 6 months for model training. The time taken to generate training data depends on the number of closed incidents on which the training data is generated. If the training data generation fails, see Troubleshooting Support group prediction issues.
Multi Attribute Prediction	Select this option to use the support group prediction and auto ticket categorization capabilities together. However, if you have been using support group prediction, you must delete or deactivate the existing support group prediction training data before generating the multi-attribute prediction training data. From the Company list, select the company for which you want to enable AI-based multi-attribute prediction. From the Application list, select Incident Management. From the Template list, select Incident Multi Attribute Prediction Training Data. Click Add Selected Template. Select the added template and click Generate Training Data. The AI algorithm analyzes closed incidents from the selected company in BMC Helix ITSM, and considers the incidents closed within the past 6 months for model training. If the training data generation fails, see Troubleshooting incident requests Note: The creation of a new Multi-Attribute Prediction data set is no longer supported. However, the existing dataset, if already created, remains usable for predictions.

Training data generation status

When you generate training data, the Data set configuration table displays the status of the training record. The status value helps you to understand the progress of your training data record.

The following table describes the status set by the system when the training data record is processed.

Status	Description
UNTRAINED	This status is set by default when you create a training data record for your company by using the out-of-the-box templates provided by BMC.
QUEUED	This status is set when a request for generating this training data record is submitted.
PROCESSING	This status is set after the training data is generated and is being used to train the cognitive service.
COMPLETED	This status is set when the system successfully trains the cognitive service by using the training data.
FAILED	This status is set when the system fails to train the cognitive service. An error message stating the reason for failure is displayed.

To deactivate a previously trained dataset

In certain situations, if you do not want to use Cognitive Service as a trained dataset, you can deactivate the option for that trained dataset to automatically categorize or automatically apply templates. For this, you must perform the following steps:

Navigate to Application Administration Console > Custom Configuration > Foundation > Advanced Options > Cognitive Service Configuration.
On the Cognitive Data Setup form, select Cognitive Data Type.
Select the training data set that you want to deactivate.
Change the Status to Inactive.