Detecting Stream Events in Distributed Streams

Events detection is one of the most important issues in data stream processing. Outlier event, change event and burst event are three typical types of event that need to be identified. In this paper, we explore the relationship of these three types of events, and then present a unified framework, named DSED (distributed stream events detection), to handle all of them simultaneously. Rather than collecting all the data in one coordinator node for centralized processing, we also use pre-computing to reduce the communication cost for events detection. Experiment results verify the efficiency of DSED, and also show that our proposed framework can dramatically reduce the communication cost with little false position.