1. Main Idea
The right decision should be comparing the probability of the scenario after getting the information. Mathematically, this probability is denoted by and . They are called a posterior probability. It means the probability of a class after knowing the data value x in a particular scenario. Therefore, the right decision is:
Decide : if
Or more general:
From the equations above, once we decide class based on the observed value x, the probability that you make a correct decision is hence . Therefore, the probability that making a wrong decision or error is
In this sense, the decision rule is optimal in the sense of minimizing thee probability of the decision error. This decision rule is called the maximum a posterior (MAP) rule.
,
which minimizes the probability of the decision error.
2. For c possible classes situation.
This is fundamental but very useful.
So, to compute , we can just add the probability of wrongly classifying to other classes. By using the formula of computing joint probability.
This equation is known as Bayesian formula.