I try to make a proof of concept of Prediction Engine, but currently it doesn’t learn anything.
I would appreciate if somebody give me some clues…
The most visible parts seems to be ok : Optimize installed, AE and PE running and looping, no nasty errors in logs, …
What I done so far :
- Created 2 KPI (KPI for measurement, KPI for the condition to predict), 2 rules and 2 eventmap associated with the fine KPI
- One prediction template targeting the rule tied to KPI to predict
- Leave PE at its default settings (15 mins for both past and future interval)
I’m able to inject any value for any KPI using a flow service and addEvent service.
I simulate big change on the KPI to measure and made fine change to KPI to predict (to trigger rule violations).
However the PE is still showing learning active, the number of KPI and rules instances stay at 0.
Is there any constraint on the rule or KPI or data history when used for prediction ?
I will shortly try to recreate the prediction template.