A Technique for Generalizing Temporal Durations in Relational Databases

We address the problem of generalizing temporal data concerning durations extracted from relational data- bases. Our approach is based on a domain generMiza- tion graph that defines a partial order specifying the generalization relations for a duration attribute. This domain generalization graph is reusable (i.e., can be used to generalize any duration attribute), adaptable {i.e., can be extended or restricted as appropriate for particular applications), and transportable (i.e., can be used with any database containing a duration at- tribute). In this paper, we propose a method for generalizing a duration attribute using domain generalization graphs; in a companion paper (14), we address the problem of generalizing calendar attributes. We define a do- main generalization graph (6) for duration attributes by explicitly identifying the domains appropriate for the relevant levels of temporal granularity and the map- pings between the values in these domains. Generaliza- tion is performed by transforming values in one domain to another, according to directed arcs in the domain generalization graph. Our goal is to specify a domain generalization graph for the duration attribute that is reusable (i.e., can be used to generalize any duration at- tribute), adaptable (i.e., can be extended or restricted as appropriate for particular applications), and trans- portable (i.e., can be used with any database containing a duration attribute). Given the duration domain gen- eralization graph and a duration attribute, our method creates a high-level map of the distinct, nontrivially dif- ferent possible summaries. Each summary is based on a different way of generalizing or a different level of time granularity. This map can be used as the basis for fur- ther automatic data mining or presented to the user, who then can choose summaries t.o examine in more detail.