I have been using R to create projects and have been reviewing the models.
If I use the platform to prioritize the highest possible model accuracy I simply choose Comprehensive mode from the Start screen.
I am not so sure how to do this using the R datarobot package. I looked at the StartProject function and used accuracyOptimizedBlueprints= TRUE to create the equivalent.
Is my thinking correct? Is there another way to maximise accuracy?
One way to test if your command has created a comprehensive project is clicking the 'Get More Accuracy' button in the UI. In Comprehensive projects the button will directs you to configure modelling settings, if it didn't run comprehensive you will see new models being made.
By default R projects modelling mode is FullAuto, defined in the docs as "represents running the entire autopilot". I'm not sure if that means comprehensive or not.
The Python API has 'COMPREHENSIVE' has a modelling mode. The R API is not as straight forward as it doesn't list comprehensive as a mode.
I ran a test myself and yes it does have "comprehensive" mode thou the documents are not clear. This method works:
projectID = "61830b14688a121f1d23ef12" SetTarget(projectID, target="Distance", mode=4)
allowed values: (0, 2, 4, 3, 'auto', 'manual', 'comprehensive', 'quick'). Hence the mode=4 for comprehensive
I haven't tried the other function calls but this at least I've tested and I can confirm it ran 84 Blue Prints in DataRobot for me