Add glossary term: Data Leakage

[HJC704]

Proposed glossary item

Data Leakage

In machine learning, data leakage occurs when information that would not be available at prediction time influences model building or evaluation, yielding overly optimistic performance estimates. In Imageomics datasets, leakage can occur when duplicate images or multiple views or records of the same biological specimen are divided between training and evaluation; keep all records with the same specimen identifier in one split.

This proposed definition is based on the Imageomics data-workshop onboarding lesson and is intended for docs/wiki-guide/Glossary-for-Imageomics.md.

References:

• Scikit-learn: data leakage
• Scikit-learn: cross-validation iterators for grouped data
• Imageomics data-workshop repository

[egrace479]

Thank you, @HJC704. Would you like to open a PR to add the term to the Glossary?