经验模态分解在数据流概要生成中的应用

Applications of EMD in generating synopses of data stream

由于流数据无限增长的特点,系统无法在内存中保存所有扫描过的流数据,因此数据流处理的关键是建立流数据的概要结构,以便随时能根据该结构提供数据流的近似处理结果,将重点讨论数据流的概要生成技术。先利用经验模态分解方法提取流数据的趋势,滤除数据中的噪声,再利用精确抽样方法实现概要的生成。利用提出的概要生成方法,内存中只需保存滑动窗口中多个段的概要信息。由于该方法中概要是基于趋势序列生成的,趋势序列较原序列平滑,序列中具有相同数值的元素增加,可以进一步节省存储空间。

Because of the infinite growth characteristic of data stream,the data stream that has been scanned can not be all saved in memory.Maintaining a synopsis data structure dynamically from data stream is vital for many streaming data appli- cations, so the paper will focus on technology to generate synopses of data stream.Firstly, use empirical mode decomposition to extract the trend of data stream, and filter out noise embedded in the data.Then use concise sampling method to generate synopses.Use synopses generation method presented in this paper, only synopses of those data, which are included in a slid- ing window, need to be saved in memory.Meanwhile, as synopses is generated based on trend sequences, which is smoother than its original sequence, so the number of the sequence＇s elements that have the same value increases, and this can further save amount of storage space.