Multi-Class Classification (MCC) Model

Edit online

The Multi-Class Classification model or the MCC model is designed to automatically categorize incidents in the Service Delivery application.

What it enables

  • Streamlined ticket triage with AI-enabled incident recommendations
  • Identifying and classifying incidents based on learning data

Defined configuration

The following configuration for Object structures, query templates, invocation channels, etc. are already configured for Maximo IT. Users can modify the query clause as per their requirement before training the MCC model.

Object Structures:

  1. Go to the Object Structures application. (Open menu > Integration > Object Structures)
  2. Open the MXAPIINCIDENTDET object. Type MXAPIINCIDENTDET in the Object Structure field and click on the record.
    Note: This is the default Object structure already created in Maximo IT.
  3. Scroll down to Source Objects for MXAPIINCIDENTDET table. The following 3 objects with corresponding values are already added:
    Table 1.
    Object Parent Object Object location path Relationship Object order User definition
    Incident   Incident   1 On
    Persongroup Incident Incident/Persongroup Ownergroup 1 On
    Longdescription Incident Incident/Longdescription Description/Longdescription 2 On
  4. Go to More Actions > Query Template.
  5. In the Query Templates for MXAPIINCIDENTDET, the following two templates are added.
    Table 2.
    Template Name Page Size User for
    INCIDENTTRAINING 1000 Training
    INCIDENTINFERENCE 10 Inferencing
  6. In the INCIDENTTRAINING template, scroll down to Query Template Attributes for INCIDENTTRAINING and the following four attributes are added:
    1. Description
    2. Description_longdescription
    3. Ownergroup
    4. Ticketuid
    Note: Description and Description_longdescrption are feature channel and Persongroup/Ownergroup are inferenced/predicted by AI for the incident.
  7. In the INCIDENTINFERENCE template, scroll down to Query Template Attributes for INCIDENTINFERENCE and the following three attributes are added:
    1. Description
    2. Description_longdescription
    3. Ticketuid
  8. Now, go to More Actions > Query Definition.
  9. In the Queries to be assignedtable, two training filter records - one for incident training and another for inference are added. The Query clause for each is also added by default. You can modify the Query clause for training and inference based on you requirement:
    Table 3.
    Query Clause Name Query Type Query Clause Is Public?
    INCIDENTTRAINFILTER osclause Ownergroup in (‘NETWORK’, ‘HARDWARE’, ‘ACCESS’, ‘ADMIN’, ‘HR’, ‘INTERNAL’, ‘MISC’, ‘PURCHASE’, ‘STORAGE’) and status = ‘NEW’ Yes
    INCIDENTINFERENCEFILTER osclause Status in (‘NEW’) and reportdate >= (CURRENT TIMESTAMP – 5 MINUTES) Yes
  10. Save the Object structure by clicking on the Save icon after modifying the Query clause for the two filters.

Invocation Channels:

  1. Go to the Invocation Channels application. (Open menu > Integration > Invocation Channels).
  2. Two channels are already created – one for training and another for inference.
  3. The first invocation channel is for AI Training for incident.
  4. To map the object structure and Query templates created in previous steps, following values are added:
    • Request object structure: MXAPIINCIDENTDET
    • Request template: INCIDENTTRAINING
    • Request Processing Class: com.ibm.tivoli.maximo.ai.AITrainReqExit
    • Response Processing Class: com.ibm.tivoli.maximo.ai.AITrainRespExit
  5. Another invocation channel for AI inference for incident is also created (AIINFINCIDENT).
  6. To map the object structure and Query templates created in previous steps, following values are already defined:
    • Request object structure: MXAPIINCIDENTDET
    • Request template: INCIDENTINFERENCE
    • Request Processing Class: com.ibm.tivoli.maximo.ai.AIInferenceReqExit
    • Response Processing Class: com.ibm.tivoli.maximo.ai.AIInferenceRespExit
  7. Save the channel.