What is a Classifier?

Classifier – generalisation of UML model elements

Classifier as a term is used in various areas. In the context of machine learning, a classifier is, for example, a special case of a hypothesis that assigns a label to a data point (e.g. when an e-mail is classified as spam or non-spam or when document properties and tags are automatically recognised and assigned). In UML, a classifier is a so-called metaclass. A metaclass is the class of a class and therefore also the basis for the instance of a class.

​As a metaclass, the Classifier is particularly useful for authors of UML – the Object Management Group (OMG) – because it can be used to describe common properties of UML model elements such as class, interface, component, attribute, method, activity, or inheritance. Each model element is thus a spezialisation of the classifier.

For UML users, a classifier is merely an intellectual construct that they cannot use in any of the seven structure diagrams

• class diagram,
• composition structure diagram,
• component diagram,
• distribution diagram,
• object diagram or
• package diagram,

or in any of the three behavior diagrams

• activity diagram,
• use case diagram or
• state (machine) diagram,

or in one of the four interaction diagrams

• sequence diagram
• interaction overview diagram,
• communication diagram or
• timing diagram.

Notes:

In the current UML 2.5 description of the OMG, you will find a whole series of classifier descriptions in chapter 7.8.