Why should group-aware splitting be used when multiple rows belong to one entity?
- A Otherwise the same entity appears in train and validation, so the model memorises rather than generalises
- B It balances the folds
- C It speeds up training
- D It is required for time series
Answer
Otherwise the same entity appears in train and validation, so the model memorises rather than generalises
Patient records, customer transactions and repeated measurements all require this, and ignoring it inflates validation scores substantially.
Log in with Google
Log in with LinkedIn
.jpg)
.jpg)
