hard20 pts · first answer countsdata-leakageclass-imbalance
SMOTE applied before cross-validation
A notebook runs SMOTE on the whole dataset and then does 5-fold cross-validation. The scores are much higher than on a later holdout. Why?
Discussion 0
Join free to post in the discussion.
Related questions
- Scaling before splittingMachine Learning Engineer
- Target encoding without leakageMachine Learning Engineer
- Several scans per patientMachine Learning Engineer
- A feature that is too goodMachine Learning Engineer
- High accuracy on rare eventsData Scientist
- Suspiciously strong churn featureData Scientist
- Using a model on rare positivesData Scientist
- Oversampling before cross-validationData Scientist