Exploring Fuzzy Ontologies in Mining Generalized Association Rules

摘要

The most common use of fuzzy taxonomies in mining generalized association rules occurs in the pre-processing stage, through the concept of extended transaction. A related problem is that extended transactions lead to the generation of huge amount of candidates and rules. Beyond that, the inclusion of ancestors may to generate redundancy problems. Besides, it is possible to see that the works have only assumed the total relation between database items and taxonomy nodes. The total relation occurs when all structure items have an equivalent representative item in the dataset, and vice-versa. Furthermore, the works have been directing for the question of mining fuzzy rules, exploring linguistic terms, but few approaches have explored new steps of the mining process. In this sense, this paper proposes the extended FOntGAR algorithm, an algorithm for mining generalized association rules under all levels of fuzzy ontologies, where the relation between database items and ontology items do not need be total. In this work the generalization is done during the post-processing step.