Customizing class imbalance handling in databricks.automl

[rtreves]

Hi Databricks! I'd like to make a feature request for end users to be able to customize class imbalance behavior in `databricks.automl`, specifically `databricks.automl.classify`. This public documentation describes the default procedure, but my team is unaware of a way for users to modify this behavior. Specifically, we'd like to be able to:

• Toggle class imbalance handling on/off
• Change the threshold at which a dataset is considered imbalanced
• Change the sampling fraction used in downsampling the major class

In the long term, availability of more complex imbalance correction techniques such as SMOTE would be appreciated too.

Thank you.

[Kaniz_Fatma]

Hi @rtreves, 

Thank you for reaching out with your feature request! Customizing class imbalance handling in databricks.automl.classify is an important consideration, especially when dealing with imbalanced datasets. 

You can submit feedback directly to the product team to influence the Databricks product roadmap in the following ways:

• To quickly submit feedback about your experience with Databricks, fill out the feedback form in your workspace.

• To interactively contribute to the product roadmap, submit a feature request in the Ideas Portal. You can view, comment, and vote up other users’ requests. You can also monitor the progress of your favourite ideas as the Databricks product team goes through their product planning and development process.