Grouping Methods for Pattern Matching in Probabilistic Data Streams

Kento Sugiura, Yoshiharu Ishikawa, Yuya Sasaki

Abstract: In recent years, complex event processing has attracted considerable interest in research and industry.Pattern matching is used to find complex events in data streams. In probabilistic data streams, however, the system may find multiple matches in a given time interval. This may result in inappropriate matches, because multiple matches may correspond to a single event. We therefore propose grouping methods of matches for probabilistic data streams, and call such merged matches a group. We describe the definitions and generation methods of groups, propose an efficient approach for calculating an occurrence probability of a group, and compare the proposed approach with a naïve one by experiment. The results demonstrate the properties and effectiveness of the proposed method.