How does Naïve Bayes relate to probabilistic sufficiency and Fisher-Neyman factorization theorem?

Naïve Bayes relies on sufficient statistics that satisfy Fisher-Neyman factorization for parameter estimation.

The Fisher-Neyman factorization theorem states that a statistic is sufficient if likelihood can be factorized into data-dependent and parameter-dependent parts. In Naïve Bayes, class counts and feature counts serve as sufficient statistics, allowing complete parameter estimation without raw data storage. This enables scalable learning and privacy-preserving computation.