How does Naïve Bayes compare to factorization machines in feature interaction modeling?

Naïve Bayes ignores feature interactions, while factorization machines explicitly model pairwise interactions.

Naïve Bayes assumes conditional independence, thus discarding feature interactions. Factorization machines extend linear models by learning latent vectors for features, enabling second-order interaction modeling. This makes FMs more expressive but computationally more complex than NB.