Visual Discovery of Patterns in Sparse Event Data Using Sentinel Event Alignment, Ranking, and Filtering

Temporal events are typically visualized on a time line indicating the date they occurred. This perspective is useful when the questions one wishes to answer are related to the date of occurrence. This view, however, makes it difficult to identify patterns related to the relative duration between events or the sequential order. We demonstrate how the simple notion of temporal alignment by some sentinel event can be a powerful method for visually identifying temporally correlated events. We put this idea into practice with the implementation of LifeTimes, a tool targeted at viewing personal event histories, such as patient medical histories. We demonstrate the power of this method on a dataset of sports injury reports.