Evidence 

The evidence variable is the probability of a condition in the dataset. In our example, the probability of temperature being 70°F or above is only 10%. Rare events have low evidence probability. Evidence probabilities boost posterior probabilities of rare events. For the purpose of the Naïve Bayes classifier, we do not need to consider the evidence variable, since it is not dependent on the class variable.

So, Bayes' theorem is used to calculate the probability of an event given a single condition. However, when we train ML algorithms, we use one or more features to predict the probability of an event. In the next section, we will explain the Naïve Bayes algorithm and see how it utilizes posterior probabilities of multiple features variables.