Elicitation of fuzzy association rules from positive and negative examples

The aim of this paper is to provide a crystal clear insight into the true semantics of the measures of support and confidence that are used to assess rule quality in fuzzy association rule mining. To achieve this, we rely on two important pillars: the identification of transactions in a database as positive or negative examples of a given association between attributes, and the correspondence between measures of support and confidence on one hand, and measures of compatibility and inclusion on the other hand. In this way we remove the “mystery” from recently suggested quality measures for fuzzy association rules.