Training data for BMC Helix Cognitive Automation
The cognitive service learns from the training data. You can use either of the following options to provide the training data.
- Data Qualifier
- CSV file
Data Qualifier
This option is the default training data type. In this data type, the training data is generated from the existing data in your system and by applying the qualifications you provide to the system. For a detailed procedure of how to generate training data by using Data Qualifier, see To generate training data by using Data Qualifier. BMC provides out-of-the-box training data templates for each application. The templates have default qualifications based on which the system generates training data.
Warning
You must use the out-of-the-box templates as is to generate training. If you try to customize the templates, the auto-categorization in Remedy with Smart IT fails.
The following table lists the out-of-the-box training data templates provided by BMC for each application.
Application | Training data template for Categorization | Training data template for ticket templates |
---|---|---|
Change Management |
| Change Template Training Data |
Incident Management |
| Incident Template Training Data |
Knowledge Management |
| Not available |
Problem Management |
| Not available |
Release Management |
| Not available |
Service Request Management |
| Not available |
Task Management System |
| Not available |
CSV file
In this data type, you need to provide a CSV file to train the cognitive service. The CSV file must contain two columns.
For auto-categorization of tickets:
- The first column must provide the description or keywords or phrases.
- The second column must provide the categories that match each description.
The records in the CSV file are used as is to train the cognitive service. You must ensure that the description text and the matching categories exist in the system. For a procedure of how to generate training data set by using CSV file, see To generate training data by using CSV file.
Sample CSV file for auto-categorization
For auto-applying ticket template to the tickets that are created using emails sent by users:
- The first column must provide the description or keywords or phrases.
- The second column must provide the unique instance ID of templates that match each description.
The records in the CSV file are used as is to train the cognitive service. You must ensure that the description text and the matching templates exist in the system. For a procedure of how to generate training data set by using CSV file, see To generate training data by using CSV file.
Training data guidelines
Understand the following guidelines before generating a training data record.
Number of training data
- You can generate a maximum of 20 training data records at a given time.
- The system throws an error if this limit is exceeded. You must delete an existing data record and then create a new record.
Data Qualifier training data
- Data Qualifier is the default training data type.
- Generate training data by using the Data Qualifier if your system has tickets that match the the qualification being used in the out-of-the-box templates and if your system has sizeable tickets with all possible categorizations.
- If the number of records in training data exceeds 15000, the system selects the most recently modified 15000 records.
CSV file training data
- Use CSV file to generate data if you have your own set of descriptions and matching categorizations or ticket templates.
- When you provide the CSV file as a training data source, ensure that it contains at least 5 records.
- If your CSV file contains more than 15000 records, only the first 15000 records are used by the system to train the cognitive service.
- If the description in the input column of the CSV file contains more than 1024 characters, only the first 1024 characters are used by the system to train the cognitive service.
Incremental training
The cognitive service is not trained incrementally. To train the cognitive service with new tickets entered in the system, you must generate new training data for the same record again.
Comments