Accelerating Sequence Event Detection through Condensed Composition

Composite event processing systems are useful in various application domains such as stock data stream monitoring, RFID data management, web access pattern monitoring, etc. The tree-based method is a typical method for implementing composite event processing system. In this paper, we introduce an optimization method called "condensed composition" for the tree-based composite event processing. "Condensed composition" divides the events in groups to participate the composition. We analyze the reason why the performance is improved and utilizes the statistics of the "condensation degree" to dynamically decide using condensed composition or not for different data characteristics. In the experiments, the "condensed composition" method significantly improved the performance on the real stock data.