Skip to main content

Generative or Naive Bayes model

Generative (Naive Bayes) model for classification is based on Bays' theorem. Here we fit a suitable probability distribution Pi(x)P_i(x) to each class of data. Then we make decision for a new point for which the probability is maximum πiPi(x)\pi_i P_i(x), where πi\pi_i is the relative frequency of ith class in the training dataset.

For the probability distribution Gaussian distribution is usual choice, but we can choose other distributions suitable to our problem.