medium10 pts · first answer countsml-pipelinesdata-leakage

A suspiciously strong feature

A loan default model has a feature "collections_calls_count" that alone gives 0.99 AUC. What is the likely problem?

A.The feature simply has too many missing valuesB.The feature needs log scaling before trainingC.It is recorded after default, so it leaks the targetD.An AUC above 0.95 is normal for loan models

Discussion 0

Join free to post in the discussion.

A suspiciously strong feature

Discussion 0

Related questions