fix smote converter

[Felipedino]

This pull request introduces improvements to the ImbalancedLearnWrapper class in the data conversion pipeline, focusing on better handling of data types and dataset splits when resampling with imbalanced-learn samplers.

Enhancements to data type management and dataset splits:

• Data type metadata is now serialized and saved into the PyArrow table using the new save_types_in_arrow_metadata utility, ensuring that type information is preserved during resampling.
• When transforming data, the output DashAIDataset now retains the original dataset splits from x, improving consistency for downstream tasks.

General improvements:

• Added a SUPERVISED class attribute to ImbalancedLearnWrapper to clarify its intended use with supervised learning workflows.

[Copilot]

Pull request overview

This pull request enhances the ImbalancedLearnWrapper class to better preserve data type metadata and dataset splits when applying resampling operations like SMOTE or ADASYN. The changes focus on serializing type information into PyArrow table metadata and maintaining dataset split information through the transformation pipeline.

Changes:

• Data type metadata is now serialized and saved into the PyArrow table using the save_types_in_arrow_metadata utility
• The transform method now preserves original dataset splits from the input
• Added SUPERVISED = True class attribute to indicate supervised learning requirement

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.