Increase accuracy of probable root cause determination

The probable root cause algorithm is currently using rule based scoring algorithm. 
Unfortunately the ranking often doesn't point to the actual root cause. One of the reasons
is the influence of the first occuring alert. Even the capability of adjusting the scoring, especially
the firstoccurence, but also other scores doesn't deliver the desired accuracy of the probable root cause.
The impact for the business is the loose of trust and confidence of the users of the tool. Another critical impact is  
the risk of disturbing the resolving in ITIL process, when creating tickets using this misleading ranking  and probable root cause determination,
which leads to increase MTTR and work. The key factor is to improve the ratio of well matched root causes to false identified root causes to decrease MTTR 
and lower effort in total significantly.

Tests and experiments by copy and paste the summaries of the alerts within an alert group and handing them to an on prem high quality LLM shows extraordinary 
good results determining the correct root cause. Together with the classic rule/score based engine this will improve the probable root cause determination 
siginficantly. Important is to use a RAG and OnPrem architecture, so that important additional internal and confidential information could be ingested and used.

Idea is: directly pass the summaries and other features of an alert group a LLM and use the result to increase accuracy of probable root cause ..