Supervised learning is about observing or directing the execution of something. The input that's given to the model is the prediction we want to make. The labeled data is the explicit prediction given for the particular instances of the input. Supervised learning requires labeled data, which requires some expertise. However, these conditions are not always met. We don't always posses the labeled dataset. For example, fraud prediction is one of the rapidly unfolding fields where the attacker is constantly looking for available exploits. These new attacks can't possibly be maintained under a dataset with labelled attacks.

Mathematically, the mapping functions of the input to the output can be expressed as Y = f(X). Here, Y is the output variable and X is the input variable.