Predictive Intelligence - Training Failed

[Martin Pavlov]

Hi there,

In the last hours I tried to resolve a problem on my own, but unfortunately could not find any solutions. I am using the PI Similarity framework, but when I click on "Submit and Train", in the Solution module it appears an error saying "Configuration OR Network Error". Could not find any information related to this error. I will appreciate any advice on how I cloud solve this issue. 

I hope this screenshots are helpful. 

Word corpus:

Solution definition:

Solution:

Thanks in advance!

[RaghavSh]

How many records do you have in cmdb_ci_outage table with task_number starting with inc and do they have some valid short description.

Try with some other table instead of cmdb_ci_outage and see the difference

[Martin Pavlov]

They have valid fields, but the records are 10. However if that was the problem I thought that there would be more descriptive error related to the error instead of "Network error". Well, after some time the error changed to "Training request timed out". I followed this guide, but the property is already populated correctly. 

And now I found this article where it says that incidents should be real. Mine are generated with script using the demo data from the dev instance. Is it possible that the problem is this?

Thanks!

[RaghavSh · Accepted solution]

you need to make sure of below points:

1. you should have 10000 valid records to train.

2.By "real incidents" servicenow means that it should have valid field values. It do works for incidents created by scripts but the field values should be accurate. Usually incidents created from scripts have similar fields values for group, short description, description, cat/subcat which cause issues in training.

You either need to create incidents with distinct values or import some real data to test this.

[Martin Pavlov]

I see, okay, thank you for your help! I will try to improve the data and will return with more information!