基于时间序列模式表示的异常检测算法  

Outlier detection algorithm based on pattern representation of time series

作　　者：詹艳艳[1] 陈晓云[1] 徐荣聪[1] 

机构地区：[1]福州大学数学与计算机科学学院,福州350002

出　　处：《计算机应用研究》2007年第11期96-99,共4页Application Research of Computers

基　　金：福建省自然科学基金资助项目(Z051503);福州大学科技发展基金资助项目(2004-XQ-17)

摘　　要：提出了一种基于时间序列的模式表示提取时间序列异常值的异常检测算法(PREOV)。时间序列的模式表示本身就具有压缩数据、保持时间序列基本形态的功能,并且具有一定的除噪能力。在时间序列模式表示的基础上提取异常值,可以大大提高算法的效率和准确性,达到事半功倍的效果。在本算法中,还使用了一定的剪枝策略,使得算法的时间复杂度进一步降低。该算法计算简单、实现方便、无须训练,可以支持时间序列的动态增长。This paper imported an algorithm which was based on the pattern representation of time series extract outlier value （PREOV）. The pattern representation of time series itself had the function of compress data and keep the basic shape of time series, and it had a certain extent effect of noises removal. Based on PREOV could enormously increase the efficiency and veracity of algorithm, and got twice the result with half the effort. Besides, used some strategy of pruning, which made the algorithm＇s time complexity more lower. And this algorithm can be easy calculated and carry out, the training is needless and it can support the dynamic increase of time series.

关 键 词：斜率 时间序列 模式表示 支持数 异常值 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]