Multi-class Classification (MCC) Model

Purpose of training

Training allows the MCC model to:

• Learn how historical service requests map to specific categories (e.g., “Network Issue,” “Access Request”)
• Recognize common patterns across descriptions and summaries
• Improve classification accuracy over time by learning from real-world data

Training Dataset Requirements

Before initiating training, ensure:

• Historical tickets are available in Maximo IT, containing fields such as summary, description, and classifications
• Classifications (i.e., target labels) are consistent and represent the real-world taxonomy used by your support teams
• A sufficient volume of labelled data exists (a few hundred to several thousand records recommended, depending on use case)

Procedure

1. Go to the AI Configuration application (Open Menu > Administration > AI Configuration).
2. Click on Add Configuration button.
3. Give a name to your Model (e.g. MCCTRAINING) and add a description.
4. In the Template field, select the template for MCC.
5. In the Template version, select a version for MCC from the list.
6. In the Object Structure field, select the object structure for AI i.e. MXAPIINCIDENTDET.
7. In the Attribute field, select the attribute you want to inference. E.g. OWNERGROUP.
8. For the Training invoke channel field, select the invocation channel created for training.
9. For the Training Filter field, select the training filter created during Query definition.
10. In the Inference invoke channel, select the invocation channel previously created for inference.
11. For the Inference Filter field, select the inferencing filter created during Query definition.
12. Click Create.
13. The model will be saved and added to the AI Configuration application page.
14. Select and open your newly created MCC model from the table in AI Configuration application.
15. Go to the Actions button and select Set arguments.
16. In the Features row, add values used for features such as description, longdescription and click Save.
17. Now click on Activate to activate the model.
18. Once activated, click Check model status and Check data requirement to run a quick data check for minimum samples passed and ensure data health.
19. Now, from Actions, click on Train model to train the model. Click on Re-train model to retrain the model.
20. Once the model is trained, click on Check model status again and it will show the latest data regarding the model such as:
    • Model ID
    • Ready – True
    • State – Ready to serve (running)
    • Last successfully trained at – Date and time of latest training
    • Trained duration
    • Model accuracy score (0-1)
21. The MCC model is now created and trained and is ready to serve.