在时间序列数据流中监测预先定义的模式,可以实现对特定应用事件的监控.本文针对原子楔形区算法存在的问题,提出双边界的概念,并在此基础上定义新的更紧密的下界距离,从而提出了更加高效的数据流过滤算法.实验证明:在不发生错误丢失的情况下,基于双边界楔形区的过滤算法比原子楔形区的方法具有更高的查询效率,对于模式间差异较大的情况,算法性能更好.
Monitoring a streaming time series for predefined patterns can accomplish monitor some special things in many applications. In view of the Atomic Wedgie algorithm's problem, this paper proposes a new more efficient streaming time series query filtering algorithm. The new algorithm introduces double wedges and defines a new more tight lower bounding distance on it. Extensive experiments demonstrate that the new algorithm can achieve tremendous improvements in the streaming time series query filtering with guaranteed no false dismissal. Especially, when the predefined patterns are more different, the algorithm is more efficient.