Predictive Intelligence – Subcategory Model Shows Precision/Recall as 0 While Category Works

[SauravR58560934]

Hello Community,

I am working with ServiceNow Predictive Intelligence (PI) to classify Incident Category and Subcategory, and I am facing an issue with training and evaluation of the Subcategory model.

What I Have Done

• Created two separate PI classification models:

  • Incident Category – Trains successfully and shows prediction results with valid Precision and Recall.

  • Incident Subcategory – Training completes, but:

    • Precision and Recall are 0, or

    • Test results are not displayed.

• Training fields used:

  • Short Description

  • Description

Issue

Even though incidents have valid subcategory values and sufficient historical data:

• The Category model works correctly.

• The Subcategory model does not show meaningful evaluation results.

Questions

1. Are Category and Subcategory dependent in Predictive Intelligence when using two different PI models?

2. Should Category be predicted first before predicting Subcategory?

3. What is the best practice for training Category and Subcategory models?

4. What are the minimum data requirements for Subcategory models (records per value, data balance, etc.)?

5. Is there any dummy/sample data available to train and test PI for Category and Subcategory, or do we need to upload our own data into the instance?

I have attached screenshots showing:

• Category model working correctly

• Subcategory model with zero or missing evaluation results

Any guidance, best practices, or documentation references would be greatly appreciated.

Thank you in advance for your help.