Fuzzy rule extraction based on the mining generalized association rules
Toshihiko Watanabe, Masanobu Nakayama · 2004
In data mining, the quantitative attributes should be appropriately dealt with as well as the Boolean attributes. This paper describes a fuzzy rule extraction method based on mining generalized association rules from database. The objectives of the method are to improve the computational time of mining and the accuracy of the extracted rules for the actual application. In our approach, we construct a hierarchical taxonomic fuzzy sets structure in each attribute. Two algorithms are shown based on the structure and the Apriori algorithm. We propose a multistage fuzzy rule extraction algorithm and a multiscan algorithm. From the results of numerical experiments, our methods are found to be effective in terms of computational time.