Question 830

Same Ratio, Every Fold

Why does stratified k-fold cross-validation specifically help when evaluating a classifier on an imbalanced dataset (e.g. 95% negative, 5% positive)?

A

It changes the loss function to penalize minority-class errors more heavily

B

It removes the need for a separate held-out test set

C

It preserves the overall class ratio in every fold, so no fold ends up missing or overweighting the minority class

D

It automatically oversamples the minority class before training

E

It guarantees every fold has an identical number of total examples