How does Naïve Bayes handle multi-label classification problems?

Naïve Bayes can be extended to multi-label classification using binary relevance or independent per-label models.

In multi-label settings, each label is treated as an independent binary classification problem. NB trains separate classifiers per label, ignoring label correlations. While simple and scalable, it fails to capture dependencies between labels, which more advanced models like classifier chains handle better.