Despite its name, the logistic regression is a classifier. As a matter of fact, the logistic regression is one of the most commonly used discriminative learning techniques because of its simplicity and its ability to leverage a large variety of optimization algorithms. The technique is used to quantify the relationship between an observed target (or expected) variable y and a set of variables x that it depends on. Once the model is created (trained), it is available to classify real-time data.

A logistic regression can be either binomial (two classes) or multinomial (three or more classes). In a binomial classification, the observed outcome is defined as {true, false}, {0, 1}, or {-1, +1}.