Abstract
A novel approach to the modeling of uncertainty in temporal information is developed by the use of Dempster-Shafer (D-S) theory. Temporal uncertain events are first represented by the use of nested intervals, using an inner more certain interval and the outer less certain interval with associated D-S mass values for each interval. The concept of the elapsed time between temporal events is defined, and detailed examples are given that illustrate the belief and plausibility for the ranges of elapsed times. The similarity for events is described where intervals' similarities are based on a relative measure of their relationships such as overlap or during. If the uncertain temporal events are determined to be similar then aggregations of such events are developed, both by averaging interval bounds and by the merging of the intervals. Finally, an initial exploration of the uncertainty for spatiotemporal data was developed by extending the nested interval uncertainty representations of our approach.