OntGAR algorithm: An ontology-based algorithm for mining generalized association rules

Most of the approaches in mining generalized association rules are focused in the extracting patterns stage, using extended transactions, and simple taxonomies. A great problem of these works is related to the generation of large amounts of candidates and rules. Beyond that, the use of taxonomies may generate some limitations like absence of formalism, problems of reuse and sharing. In this sense, this paper proposes a new algorithm for mining generalized association rules. The originality of this work is on the fact of the generalization being done in the post-processing stage and under all levels of ontologies, which are structures used in a formal domain specification. Some relevant points are the specification of a new methodology of generalization, including a new method of grouping rules; and a new and efficient method for calculating both the support and confidence of the generalized rules.