How does Naïve Bayes behave under extreme class imbalance?

Naïve Bayes becomes biased toward the majority class unless priors or weights are adjusted.

In imbalanced datasets, prior probabilities dominate posterior computation. If not corrected, the model may always predict the majority class. Techniques like prior adjustment, resampling, or cost-sensitive learning help mitigate this issue. However, NB can still perform surprisingly well in skewed text classification tasks.