EventScope: Finding Discriminating Event Patterns for Different Outcomes

Discovering temporal patterns in large datasets is a challenging problem because researchers must explore a huge space of possible correlations. EventScope is a visualization and ranking tool to assist in identification of temporal patterns leading to, or following, different events. In order to identify non-trivial patterns, EventScope splits the data based on user-specified reference events. Patterns are sorted based on differences found between these two groups. By providing information regarding an event of interest relative to two outcomes, more useful temporal patterns can be identified. EventScope was evaluated by conducting usability tests with events derived from one season of Chicago Bulls basketball games. Usability test results from non-expert users were generally positive and users were able to successfully identify several strong relations.