Identifying and Measuring Associations of Temporal Events

Large databases of temporal records have made it possible for researchers to verify their hypotheses related to temporal event sequences. However, with the overwhelming size of data and numerous possible patterns, an important issue is what patterns should be highlighted and presented to users. We implement a visualization tool, PairFinder, to enable users to efficiently locate patterns of interest. Users can 1) see all the results of the potential event patterns and 2) use interestingness measures to rank event patterns by their interestingness. In addition, users can hide irrelevant patterns and filter records by record attributes. By looking only at the topranked patterns, users can easily scan large number of patterns. We demonstrate the potential of PairFinder with four case studies and summarize the patterns found in the data sets. ∗e-mail: [email protected] †e-mail:[email protected] ‡e-mail:[email protected]