Predictive Intelligence - Clustering -> How to show on a record level?
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
05-10-2022 05:20 AM
I have built a Performance Intelligence clustering model -> Solution definition on the Incident table.
And it also has built an ML solution. Now if a new incident record is inserted in the table and a train is updated and added to the cluster.
How would we get to know or will there be any change on the record level?
- Labels:
-
Predictive Intelligence

- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
05-10-2022 08:49 AM
Not sure if I completely understand your question, so please clarify if I am interpreting incorrectly.
If you set the update frequency (see fig1) for the cluster each time a new incident is created it will be added to the existing clusters, in this case we will add new records to each cluster every 15 minutes. The training frequency will re-train the entire model and create NEW clusters, in this case we will retrain every 30days.
(fig 1 - update frequency)
The new records should be reflected in record count that you see in the below tree map.
HTH - Lener
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
05-10-2022 10:37 PM
That's correct, the incident will attach to the cluster.
But how's it going to be helpful for me to resolve the incident? Will, there be any change on the incident record. Or it is just adding in the clustering for new incident records and to visualise the cluster to automate the process.

- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
05-15-2022 10:31 PM
Clustering is used to identify patterns for a group of records. For example you can run cluster analysis on incident short descriptions to find potential automation opportunities, or run clustering on root cause to find patterns for root cause for particular incidents. It doesn't do anything to the incident records OOTB.
If you are trying to use Predictive Intelligence to help you resolve the incident then you will want to look at leveraging similarity, which can help find similar resolved incidents w/ potential work notes that you can use to solve the incident you are working on. You can reference the K21 Accelerate Incident Resolution for Agents using AI/ML and Flow Designer class to better understand how to leverage similarity to help you resolve incidents.