Stacked Ensemble not winning in AutoML

Description

Here are some examples where the Stacked Ensemble is not winning the leaderboard in AutoML. We should investigate what’s happening here and see if we can improve the results by adjusting the metalearner.

 

  • The regression example here with 80 models.

  • A binary classification example here but the AUC is very high (0.999) so maybe this is not that big of a concern since that’s common.

 

More examples are summarized in this private notebook:

Assignee

New H2O Bugs

Fix versions

None

Reporter

Erin LeDell

Support ticket URL

None

Labels

None

Affected Spark version

None

Customer Request Type

None

Task progress

None

CustomerVisible

No

Components

Priority

Major
Configure