A dynamic burst detection model over data streams

Abstract

Burst detection techniques over data streams has been attracting board and home scholars' more attention due to it's broad applications in financial, medical service, telecommunication and other critical important areas. In order to detect bursts of positive data streams, negative data streams, first, we propose a dynamic burst detection model over data stream. Based on the model, we embed a two-dimensional array into SAT(shifted aggregation tree), and construct a elastic data structure ASAT given the input. At last, we propose a elastic burst detection algorithm over data streams. The algorithm not only can detect bursts of monotonous accumulation function and non-monotonous accumulation function, but also can search burst on large scale of negative, constant data streams. Experiments show that this algorithm is both efficient and effective.