Is DataRobot a Black Box?

DataRobot—through its blueprints and documentation—provides complete transparency for how models are built. And blueprints show the preprocessing steps, algorithms, and all associated parameters. Complete documentation and references to any applicable academic literature are linked from each blueprint.

DataRobot provides missing values reports and full compliance reports for model risk teams.

Additionally, DataRobot offers a suite of interpretability tools to help you understand your model. Feature Impact identifies which features are driving your model, while the Feature Effects tools use partial dependence to show how a specific feature affects the target. Finally, Prediction Explanations work at the level of individual examples to show exactly how the model is making a decision.

All of these work together to ensure you can reproduce your model as well as understand how it is working.

More Information

  • See the Artificial Intelligence Wiki: Prediction Explanations and Feature Impact.
  • If you’re a licensed DataRobot customer, search the in-app documentation for Blueprints, Model Compliance, and Understand division.
Version history
Revision #:
15 of 15
Last update:
‎05-15-2020 04:20 PM
Updated by:
 
Contributors