一种流量无关的数据流突发检测算法

针对数据流整体数据量波动剧烈的情况下准确突发检测问题，改进了经典的聚合塔数据结构，提出了一种流量无关的数据流突发检测方法FFBD(flow free burst detection)。该方法使用单个元素数量与总体元素数量的比值作为单个元素滑动窗口内的聚合函数值，并使用前后2个滑动窗口内的聚合函数的比值来判断突发，该数据结构能够高效支持弹性窗口时间跨度不大的应用场景。实验结果证明，与典型的聚合塔突发检测方法相比，FFBD方法虽然多使用了2%的存储空间、提高了5%的计算复杂度，但能有效地规避整体数据流的显著变化对单个元素突发检测带来的影响，有着更好的检测效果。

Flow free burst detection algorithm on data stream

A Flow Free Burst Detection method called FFBD was presented based on classical aggregation pyramid data structure to address the accurate burst detection problem when the total flow of data stream fluctuates acutely. This method makes use of the ratio of one element's frequency and the total elements' frequency as the aggregation fuction value to check burst. The algorithm can support the scene where the elastic sliding window is not so big. Experiment show that compared with the traditional aggregation pyramid data structure, FFBD provides the capability of avoiding the influence caused by the total flow at the cost of 2% more space and 5%more computational complexity.