Mining Fuzzy Association Rules on Has-A and Is-A Hierarchical Structures

Preliminary studies on data mining focus on finding association rules from transaction databases containing items without relationships among them. However, relationships among items often exist in real applications. Most of the previous works only concern about Is-A hierarchy. In this paper, hierarchical relationships include a Has-A hierarchy and multiple Is-A hierarchies are discussed. The proposed method first reduces a Has-A & Is-A hierarchy into an extended Has-A hierarchy using the IsA-Reduce algorithm. The quantitative data is transformed into fuzzy items. The RPFApriori algorithm is then applied to find fuzzy association rules from the fuzzy item data and the extended Has-A hierarchy.