[1]刘雪梅,王亚茹.基于异常因子的时间序列异常模式检测[J].计算机技术与发展,2018,28(03):93-96.[doi:10.3969/ j. issn.1673-629X.2018.03.019]
　LIU Xue-mei,WANG Ya-ru.Anomaly Pattern Detection in Time Series Based on Outlier Factor[J].,2018,28(03):93-96.[doi:10.3969/ j. issn.1673-629X.2018.03.019]
点击复制

基于异常因子的时间序列异常模式检测()
分享到：

《计算机技术与发展》[ISSN:1006-6977/CN:61-1281/TN]

卷:
28
期数:
2018年03期
页码:
93-96
栏目:
智能、算法、系统工程
出版日期:
2018-03-10

文章信息/Info

Title:
Anomaly Pattern Detection in Time Series Based on Outlier Factor
文章编号:
1673-629X(2018)03-0093-04
作者:
刘雪梅; 王亚茹
华北水利水电大学 信息工程学院,河南 郑州 450045
Author(s):
LIU Xue-mei; WANG Ya-ru
School of Information Engineering,North China University of Water Resources and
Electric Power,Zhengzhou 450045,China
关键词:
关键词:时间序列; 分段线性表示; 异常模式; 异常因子; 子序列; 特征空间
Keywords:
time series; piecewise linear representation; anomaly pattern; outlier factor; subsequence; feature space
分类号:
TP31
DOI:
10.3969/ j. issn.1673-629X.2018.03.019
文献标志码:
A
摘要:
时间序列中的异常模式能够提供大量有意义的信息,由于时间序列数据量大、含噪音、维度高,直接在原始时间序列数据中进行异常模式挖掘要花费大量的时间和空间代价。 常用的时间序列分段线性表示法,易受阈值和分段数目的影响。 对此,根据实际工程监测中时间序列的特征,将不限定分段数目与子序列长度的方法相结合,基于斜率及最大时间跨度,将原始时间序列分割成长度不同的子序列,提取子序列的极值差、斜率、均值等特征值,并映射到三维特征空间,在该特征空间中计算正常模式间的距离,以正常模式间距离为标准,求出各子序列的异常因子,检测异常模式。 为验证该算法的有效性,采用南水北调工程安全监测中的实测数据和人工合成数据进行测试,取得了较好的效果。
Abstract:
Anomaly patterns of time series can provide a lot of meaningful information. Because of the large amount of data,noises and high dimension for time series,anomaly pattern mining in the original time series directly will take much time and space. The commonly used piecewise linear representation methods are vulnerable to the threshold and the number of segments. For this,based on the time series in engineering monitoring,we combine the method of not limiting the number of segments and the length of the subsequence and segment the time series based on slope and time span. Then the extreme difference,slope and mean value of these sections are extracted and transformed into the three-dimensional feature space,where the distance of the normal pattern is calculated as the standard to solve the outlier factors of each subsequence for detection of anomaly patterns. We demonstrate the effectiveness of the proposed algorithm through an application to an actual dataset from South-to-North Water Transfer Project as well as an artificial dataset,with better results.

相似文献/References:

[1]李云霞,姚建国,万定生,等.一种水文时间序列异常模式检测方法研究[J].计算机技术与发展,2019,29(07):159.[doi:10. 3969 / j. issn. 1673-629X. 2019. 07. 032]
　LI Yun-xia,YAO Jian-guo,WAN Ding-sheng,et al.An Anomaly Pattern Detection Method for Hydrological Time Series[J].,2019,29(03):159.[doi:10. 3969 / j. issn. 1673-629X. 2019. 07. 032]
[2]刘学彬,梁智飞,朱卫平,等.基于 EEMD 的固定分段数分段线性表示方法[J].计算机技术与发展,2023,33(11):202.[doi:10. 3969 / j. issn. 1673-629X. 2023. 11. 030]
　LIU Xue-bin,LIANG Zhi-fei,ZHU Wei-ping,et al.Piecewise Linear Representation Algorithm of Fixed Section Number Based on EEMD[J].,2023,33(03):202.[doi:10. 3969 / j. issn. 1673-629X. 2023. 11. 030]

更新日期/Last Update: 2018-04-25