Predictive intelligence clustering is throwing error while training in PDI?

[Community Alums]

Hi all,

when I am trying to train the OOB std change template candidates clustering definition , first the training state is in Unauthorized and after sometimes it's throwing error Training cancelled and Training timeout why this is happening in my PDI. Training is not getting success.

What is the user of this OOB cluster std change template candidates  for  Standard Change Predictive Intelligence  for Change

We can able train the cluster in PDI?

can anyone help me this

Thanks

[davitvoski]

Amazing. Our instance was missing step 1. Thank you very much Brain. 

Now we can provide the best PI out there!😀
Is this documented in any of the documentation by any chance?

[Brian Bakker]

@davitvoski 

The system property is documented in Predictive Intelligence references, however, it does not state the value it should be set to, so I will get this updated.

If the solution training throws an error at 75%, then the issue is with the dataset where data is skewed towards one class or contains too many empty values. Ensure you solution definition filter has the condition IS NOT EMPTY for your Input/Output fields, and for Classification solutions, you GROUP BY the Output field to see how many records are in the dataset for each class. Less than 30 records, and the class will be excluded. More than 30 records does not guarantee it can make predictions on the class, as it is dependent on the quality of the data in the records for the class. For classes that are generated where it cannot make predictions, it will set Precision=100, Coverage=0, Recall=0, which will set the Threshold=100.

If 80% of the records in the training dataset is skewed towards one class, it may also fail with "Error while training solution". It will throw the following exception in the localhost logs -

2024-01-31 01:40:30 (164) Default-thread-1 6821448A8708869400C210AABBBB357C txid=ac21448a7c86 Training solution 2cd0048a8708869400c210aabbbb3511, stage: Training Solution, stage progress: 100, details: SE0060:Solution Training Failed. Ask Support to use log key 691777 to investigate trainer logs further., logging details: java.lang.NullPointerException
at com.snc.ml.api.impl.DxCSolution.isPredictColumnSkew(DxCSolution.java:426)

Always check the underlying dataset.

Finally, @Lener Pacania1's Community Articles on Predictive Intelligence (PI) are their weight in gold and provide excellent guidance on implementing PI.

Wishing you success with your PI implementations.

Regards,
Brian

[davitvoski]

Hey Brian,

Do you know anyone you can connect me with for AI Search SharePoint Integration?

Thank you

[Brian Bakker]

@davitvoski 

You posted this question in the wrong topic. Please create a new question in the correct topic, so that I can answer. Thank you.

[E Chow]

Lener: For a feature that is being spruiked and the acceleration of GenAI, does it continue to be ServiceNow's position that you won't support PDI training? I see updates are constantly being deployed for the PDI. For example today i just updated the ml_sn_global resolved and open cases so I'll see if i can use it.

but I am still receiving training failed for Project name similarity. see image.