midNaïve Bayes
What is prior probability in Naïve Bayes and how is it estimated?
Updated May 17, 2026
Short answer
Prior probability represents the probability of a class before observing any features.
Deep explanation
Prior probability P(C) is estimated as frequency of class C in training data. It encodes baseline belief about class distribution. In imbalanced datasets, priors significantly influence predictions and may require reweighting or correction.
Real-world example
In spam detection, spam emails may form 20% of dataset, so P(spam)=0.2.
Common mistakes
- Ignoring class imbalance leading to biased predictions.
Follow-up questions
- How does class imbalance affect Naïve Bayes?
- Can priors be manually set?