Set the modeling mode¶
In this tutorial, you'll learn how to choose the appropriate modeling mode from the four options offered by DataRobot: Autopilot, Quick (the default), Manual, and Comprehensive. A project's modeling mode controls the sample size percentage of the training set that DataRobot uses to build models.
Takeaways¶
This tutorial explains:
- How to select a modeling mode
- Differences between modeling modes
- When to use each modeling mode
Select the modeling mode¶
After selecting a target feature for your project, the ability to choose a modeling mode becomes available. To set the project's modeling mode, click the Modeling Mode dropdown.
Use the section below to determine which modeling mode is the best fit for your use case.
Autopilot¶
In full Autopilot, DataRobot starts by building models using 16% of the total data. The 16 models with the highest accuracy score move on to the next round of modeling where they are rerun on 32% of the data. In round 2, the top 8 most accurate models move on to another round of modeling where they are rerun on 64% of the data (or 500MB of data, whichever is smaller).
By default, Autopilot runs on the Informative Features feature list.
When is Autopilot useful?
When you want to generate a more diverse group of models for your specific use case (results in slower runtimes).
Quick¶
Quick (Autopilot)—the default modeling mode in DataRobot—is a shortened and optimized version of full Autopilot mode, running on 64% of the data. It builds a smaller group of models, whose selection is based on the target feature and performance metric. The intent is to quickly generate a set of models.
When is Quick mode useful?
When you want to generate the best models for your specific use case using a balance of speed and accuracy.
Manual¶
Manual mode allows you to select and run individual models from the model Repository.
When is Manual mode useful?
When you already know which kind of modeling approach you want to use.
Comprehensive¶
Comprehensive mode allows you to run all blueprints in the Model Repository, using the maximum Autopilot sample size to ensure more accurate models.
When is Comprehensive mode useful?
When you want to find the most accurate model for your use case, regardless of time. Note that this mode can result in significantly longer build times.
Learn more¶
Documentation:
- Descriptions of modeling modes
- Circumstantial modeling mode behaviour
- Automatically created feature lists