Supervised learning is the key concept behind such amazing things as voice recognition, email spam filtering, and face recognition in photos, and detecting credit card frauds. More formally, given a set, D, of learning examples described with features, X, the goal of supervised learning is to find a function that predicts a target variable, Y. The function, f ,that describes the relation between features X and class Y is called a model:

The general structure of supervised learning algorithms is defined by the following decisions (Hand et al., 2001):

1. Define the task
2. Decide on the machine learning algorithm, which introduces specific inductive bias; that is, and a priori assumptions that it makes regarding the target concept
3. Decide on the score or cost function, for instance, information gain, root mean square error, and so on
4. Decide on the optimization/search method to optimize the score function
5. Find a function that describes the relation between X and Y

Many decisions are already made for us by the type of the task and dataset that we have. In the following sections, we will take a closer look at the classification and regression methods and the corresponding score functions.