Our use case:
We would like to allow our end-users of the Celonis analyses to test hypotheses themselves (does attribute X correlate with KPI y?).
As we are currently still on-premise we are investigating connecting Celonis to an R-Server to enable the RCALL function. This would allows us to dynamically build the data set to pass to R and then present the result in a user-friendly way.
However, as we might move to the IBC in the (near or far) future, we wondering if such analysis <-> ML interaction is also possible with the ML Workbench. Concretely, can we pass parameters from the analysis to a stored ML function/notebook which then returns the results to be presented in the analysis?
An alternative would be to pre-calculate every possible combination of attribute (sets) and KPIs, which I guess is not handy.
Looking forward to some inspiration!