The development of systems for knowledge discovery in databases, including the use of association rules, has become a major research issue in recent years. Although initially motivated by the desire to analyse large retail transaction databases, the general utility of association rules makes them applicable to a wide range of different learning tasks. However, association rules do not accommodate the temporal relationships that may be intrinsically important within some application domains. In this paper, we present an extension to association rules to accommodate temporal semantics. By finding associated items first and then looking for temporal relationships between them, it is possible to incorporate potentially valuable temporal semantics. Our approach to temporal reasoning accommodates both point-based and interval-based models of time simultaneously. In addition, the use of a generalized taxonomy of temporal relationships supports the generalization of temporal relationships and their specification at different levels of abstraction. This approach also facilitates the possibility of reasoning with incomplete or missing information.
展开▼
