cancel
Showing results for 
Search instead for 
Did you mean: 

Can we re-create DataRobot sklearn models in Python sklearn?

Jishi
Blue LED

Can we re-create DataRobot sklearn models in Python sklearn?

I was trying to recreate model built in DataRobot (Adaboost Regressor) in local python sklearn environment, with same model hyper-parameters.  But I'm getting different prediction results.

Any help?

1 Reply
David Moxley
Data Scientist
Data Scientist

In short, DataRobot guarantees internal reproducibility (i.e. same data, same settings, same outputs within DataRobot), but not external reproducibility - recreating a DataRobot Blueprint outside of the platform using the same hyperparameters as you're suggesting. There are number of factors to consider as to why external reproducibility is difficult to achieve beyond matching model hyperparameters to include data partitioning and the specific implementation of common algorithms by DataRobot within the application.

If your objective is to deploy your model outside of DataRobot, there are several ways you can leverage your chosen DataRobot model outside of the platform to score new data including Scoring Code and DataRobot Prime models.