On Co-occurrence Pattern Discovery from Spatio-temporal Event Stream

The proliferation of location-acquisition technologies and online social networks such as twitter, Foursquare, Meetup lead to huge volumes of spatio-temporal events in the form of event stream. In this study, we investigate the problem of discovering spatio-temporal co-occurrence patterns from spatio-temporal event stream (CoPES). We propose an effective sliding-window based dynamic incremental and decayed (abbreviated as DIAD) algorithm for discovering CoPES. DIAD algorithm proposes a novel decay mechanism to calculate the prevalence of CoPES and a sliding-window to process the event stream time slot by time slot to discover CoPES. The algorithm utilizes a hash tree to store the closet COPES. Then the decay mechanism and the sliding-window exploit the superimposed spatio-temporal neighbor relationships between time slots to get the accurate prevalence from event stream and discover CoPES efficiently. The experimental results on real dataset show that our proposed algorithm has superior quality and excellent expansibility.