改进的基于移动小波树的数据流异常检测方法

Improved algorithm of anomaly detection in data streams based on Shifted Wavelet Tree

针对移动小波树数据流异常检测算法的不足,提出了一种改进的移动小波树异常检测方法,利用比率阈值去除颠簸数据的干扰,提高了检测精度;利用二分查找检测算法,提高了检测的效率;结合实时增量更新算法满足了数据流在线处理的要求;改进了阈值设定方法,可实现双边异常检测。用射线数据和电能质量扰动数据进行仿真实验,结果验证了方法的有效性。

Abstract： Aiming at the limitation of the anomaly detection algorithm of shifted wavelet tree in data streams, this paper proses an improved algorithm which enhances the accuracy of detection for eliminating the disturbance of bumps using ratio threshold, increases the efficiency of detection by the means of binary search, meets the requirement of the online processing of the data streams combing with the real - time incremental updating algorithm, realizes the double sizes detection by improving the method for setting threshold. The simulation tests are carried out using the data of Gamma Ray and Power Quality Disturbance. The results validate the effectiveness of the method.