地震监测时间序列异常值检测策略
|
摘要
前兆观测数据真实可靠地应用在地震预测中前提是观测人员能够准确去除和标记干扰异常数据。为了能够在前兆数据预处理过程中由计算机及时检测出异常值,文章在时间序列的异常值检测的基础上,提出一套前兆仪器数据异常判定方法。根据异常的特征分析,首先选取合适的阈值特征进行阈值异常检测,在阈值异常检测结果的基础上,进而将其与邻近观测点数据进行相似性度量。从而达到将局部干扰异常与地球物理变化区分的目的。这种异常判定规则应用在前兆预处理中大大提高了前兆数据预处理的准确性和效率。同时,这种异常判定规则应用在实时监控中可以及时发现仪器的故障以及解决异常数据隐蔽性和时效性的问题。
Pretreatment able to accurately remove Abnormal interference data is the premise of Precursor observation data applied in earthquake prediction reliably.in order to detect outliers in the process of precursory data pretreatment,based on the time series outlier detection,put forward a set of method for judging abnormal According to the analysis of abnormal characteristics,abnormality judging rules are divided into user defined threshold rules and similarity measure rule.and then were given two types of abnormal decision algorithm.First of all,Select suitable threshold characteristics for anomaly detection,on the basis of the result of threshold outlier detection,Then,process similarity measurement with the data of the adjacent observation point.So as to achieve the purpose of making a distinction between local disturbance and geophysical changes.
Pretreatment able to accurately remove Abnormal interference data is the premise of Precursor observation data applied in earthquake prediction reliably.in order to detect outliers in the process of precursory data pretreatment,based on the time series outlier detection,put forward a set of method for judging abnormal According to the analysis of abnormal characteristics,abnormality judging rules are divided into user defined threshold rules and similarity measure rule.and then were given two types of abnormal decision algorithm.First of all,Select suitable threshold characteristics for anomaly detection,on the basis of the result of threshold outlier detection,Then,process similarity measurement with the data of the adjacent observation point.So as to achieve the purpose of making a distinction between local disturbance and geophysical changes.
引文
[1]V.Barnett,T.Lewis.Outliers in statistical data.John Wiley and sons,1994.
[2]成万里,熊豪,曲翠兰.前兆仪器数据异常实时监控系统研究[J].数字技术与应用,2012(10):12-14.
[3]翁颖钧,朱仲英.基于动态时间弯曲的时序数据聚类算法的研究[J].计算机仿真,2004,21(3):37-40.
[4]董晓莉,顾成奎,王正欧.基于形态的时间序列相似性度量研究[J].电子与信息学报,2007,29(5):1228-1231.
[5]Edward Hung,David W.Cheung.”Parallel Mining of Outliers in Large Database”,in Distributed and Parallel Database(DAPD),Kluwer Academic Publishers,Volume 12,Issue 1,pages 5-26,July2002.
[6]郑健,皮德常.基于共享最邻近的聚类和孤立点检测算法.第一届中国高校通信类院系学术研讨会,2007.
[7]陈英,顾国昌,吕天阳.基于离群点识别的聚类结果属性特征簇发现.哈尔滨工程大学学报,3/2009,3(30).
[2]成万里,熊豪,曲翠兰.前兆仪器数据异常实时监控系统研究[J].数字技术与应用,2012(10):12-14.
[3]翁颖钧,朱仲英.基于动态时间弯曲的时序数据聚类算法的研究[J].计算机仿真,2004,21(3):37-40.
[4]董晓莉,顾成奎,王正欧.基于形态的时间序列相似性度量研究[J].电子与信息学报,2007,29(5):1228-1231.
[5]Edward Hung,David W.Cheung.”Parallel Mining of Outliers in Large Database”,in Distributed and Parallel Database(DAPD),Kluwer Academic Publishers,Volume 12,Issue 1,pages 5-26,July2002.
[6]郑健,皮德常.基于共享最邻近的聚类和孤立点检测算法.第一届中国高校通信类院系学术研讨会,2007.
[7]陈英,顾国昌,吕天阳.基于离群点识别的聚类结果属性特征簇发现.哈尔滨工程大学学报,3/2009,3(30).
版权所有：© 2023 中国地质图书馆 中国地质调查局地学文献中心