Constraining Forecast limits

[adsch1]

Hi all,

Was wondering if there is a way to constrain your forecast to never exceed (or go below) a certain value. Like for instance, if you know your series can never go below 0, yet a model is forecasting negative values, can you adjust this in some setting or transformation? Thank you

[desmond_lim]

@adsch1 yes my colleague is referring the DataRobot MLOps capability to constrain predictions within limits and that is covered in our MLOps classes of which the first is listed below:

https://university.datarobot.com/mlops-i

However, if it is giving you unexpected predictions perhaps there may be changes in the data and outcomes that may require attention/retraining.

Or perhaps you may just wish to normalize your data to remove the negative values.

[senkin13]

Hi adsch1,

I think Humility rules can help you to constrain forecast limits.Please read the document blew

https://docs.datarobot.com/en/docs/mlops/governance/humble.html