Grouped split leakage audit
Locate grouping keys assigned to more than one evaluation fold.
Your entries stay in this page’s memory and are not submitted to us. Downloads are generated locally. Reloading or leaving the page clears your entries. Do not enter patient information.
The starting values are hypothetical. Use aggregate or synthetic inputs only.
Calculation results
- Leaking groups
- 1
- Conflicting assignments
- G1 (train, test)
- Distinct groups
- 3
How to interpret the output
Use a grouping level appropriate to the deployment question, such as site or synthetic subject group. Multiple records within one fold are allowed. Disjoint groups do not prove that feature engineering, tuning or outcome timing avoided leakage.
Method reference