Association rule evaluation for classification purposes
Fernando Berzal, Juan Carlos Cubero Talavera, Nicolás Marı́n, José-Marı́a Serrano, Daniel Sánchez Fernández, M.A. Vila · 2005
Association rules [1] [2] add statistical measures, such as support and confidence, to the classic rules in Propositional Logic. In order to make that difference obvious, their representation uses ⇒ instead of the classic →. An association rule is an implication X⇒Y where X and Y are itemsets with empty intersection (i.e. sets with no items in common). The intuitive meaning of such a rule is that when X appears, Y also tends to appear. The confidence of an association rule X⇒Y is the proportion of the transactions containing X which also contain Y. The support of the rule is the fraction of the database which contains both X and Y. Classification models can be directly built from traditional association rules, using association rules where the class appears in their right-hand side. In the following Section, we describe some of the approaches followed to build classification models from association rules. Afterwards, we will analyze the use of alternative measures to evaluate the suitability of association rules for classification models. Such a study is important in order to choose the best association rules what will become part of the classification model.