We have multiple cases when we want to make sure there is an explicit correlation between two data series apart from visually representing those series on a graph. Is there any simple KPI/PQL formula we can use, or do we need to go with a workaround deriving corr from calculating standard deviations etc?
Question
Is there any way to calculate the correlation coefficient with PQL?
+4
+26
A simple solution for this is not available as far as I'm aware of. What you could do, is using the Machine Learning capabilities and load the data into statistical packages. I'm doubting however if this is easier that just calculating the values yourselves with the help of the mathematical functions.
If you do want to use the progamming languages, you could follow this:
- For the Cloud solution, you could set-up a Python script in the Machine Learning workbench that extracts the serie, calculates the statistics you want, and let it return in a KPI component. In the analysis you could use the 'Run ML-learning script' component to trigger the calculation.
- If you are running a on-premise version, you could establish this via the R-integration, which works directly from a PQL statement (this is a lot easier, would love to see this integration also in the Cloud).
Kind regards,
Jan-peter
you might consider exploring custom KPIs in Celonis to approximate correlation. You can compare process measures and see trends that way. Has anyone tried this method, or perhaps created templates for it? Would love to hear how others handle similar analyses using PQL!
Login to the community
No account yet? Create an account
Login with SSO
Celonaut Login Celonis ID LoginEnter your E-mail address. We'll send you an e-mail with instructions to reset your password.