基于GPS数据的震前短临异常检测  

Short impending anomaly detection before earthquake based on GPS data

作　　者：李南[1] 林莉莉 LI Nan;LIN Lili(College of Computer and Information Sciences,Fujian Agriculture and Forestry University,Fuzhou 350000,China)

机构地区：[1]福建农林大学计算机与信息学院,福州350000

出　　处：《智能计算机与应用》2022年第7期59-68,共10页Intelligent Computer and Applications

基　　金：福建省中青年教师教育科研项目(JAT190142);福建省自然科学基金(2019J05048)。

摘　　要：随着GPS台站的普及,结合GPS数据进行时间序列数据异常检测已成为热门研究领域。针对现有方法普遍存在的主观性强、普适性差等问题,运用鞅理论,提出了一种基于GPS数据的震前短临异常检测算法(Anomaly Detection Algorithm based on GPS data, ADA)。实验结果表明,ADA算法所检测到的GPS数据中,异常出现时间与地震发生时间存在显著相关,与时间序列异常检测中传统的kσ准则和主流的异常检测模型ARIMA、单类别支持向量机OCSVM以及基于两阶段聚类的异常检测算法TSOD相比,ADA算法能够更直观、准确地反映震前GPS数据中出现的异常,不易出现误报的情况。Short impending anomaly detection before earthquake is a key part of the earthquake early warning. With the rapid popularization of GPS stations, anomaly detection before earthquake based on GPS data has become a hot research area. In order to solve the problem of strong subjectivity and poor universality in the existing methods, an algorithm based on Martingale theory is proposed to detect the short impending anomalies in GPS data. The experimental results show that the detected short impending anomaly from GPS data has a significant correlation with the corresponding earthquake. Compared with the traditional kσ analysis method, the famous anomaly detection model ARIMA, one-class support vector machines(OCSVM) and two-stage clustering algorithm for outlier detection(TSOD), the proposed algorithm can reveal the pregnancy of the earthquake more clearly and accurately, and has less false alarms.

关 键 词：地震 短临异常 GPS 鞅理论 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]