User - De ned Association

Discovering interesting associations of events is an important data mining task. In many real applications, the notion of association, which deenes how events are associated, often depends on the particular application and user requirements. This motivates the need for a general framework that allows the user to specify the notion of association of his/her own choices. In this paper we present such a framework, called the UDA mining (User-Deened Association Mining). The approach is to deene a language for specifying a broad class of associations and yet eecient to be implemented. We show that (1) existing notions of association mining are instances of the UDA mining, and (2) many new ad-hoc association mining tasks can be deened in the UDA mining framework.