Anomaly detection of earthquake precursor data using long short-term memory networks

Yin Cai, Mei Ling Shyu, Yue Xuan Tu, Yun Tian Teng, Xing Xing Hu

Research output: Contribution to journalArticle

Abstract

Earthquake precursor data have been used as an important basis for earthquake prediction. In this study, a recurrent neural network (RNN) architecture with long short-term memory (LSTM) units is utilized to develop a predictive model for normal data. Furthermore, the prediction errors from the predictive models are used to indicate normal or abnormal behavior. An additional advantage of using the LSTM networks is that the earthquake precursor data can be directly fed into the network without any elaborate preprocessing as required by other approaches. Furthermore, no prior information on abnormal data is needed by these networks as they are trained only using normal data. Experiments using three groups of real data were conducted to compare the anomaly detection results of the proposed method with those of manual recognition. The comparison results indicated that the proposed LSTM network achieves promising results and is viable for detecting anomalies in earthquake precursor data.

Original languageEnglish (US)
JournalApplied Geophysics
DOIs
StateAccepted/In press - Jan 1 2019

Fingerprint

earthquake precursor
earthquakes
anomalies
anomaly
preprocessing
predictions
earthquake prediction
detection
prediction

Keywords

  • anomaly detection
  • deep learning
  • Earthquake precursor data
  • LSTM-RNN
  • prediction model

ASJC Scopus subject areas

  • Geophysics

Cite this

Anomaly detection of earthquake precursor data using long short-term memory networks. / Cai, Yin; Shyu, Mei Ling; Tu, Yue Xuan; Teng, Yun Tian; Hu, Xing Xing.

In: Applied Geophysics, 01.01.2019.

Research output: Contribution to journalArticle

@article{9e46748afa42488f9cbb7a2e8241460e,
title = "Anomaly detection of earthquake precursor data using long short-term memory networks",
abstract = "Earthquake precursor data have been used as an important basis for earthquake prediction. In this study, a recurrent neural network (RNN) architecture with long short-term memory (LSTM) units is utilized to develop a predictive model for normal data. Furthermore, the prediction errors from the predictive models are used to indicate normal or abnormal behavior. An additional advantage of using the LSTM networks is that the earthquake precursor data can be directly fed into the network without any elaborate preprocessing as required by other approaches. Furthermore, no prior information on abnormal data is needed by these networks as they are trained only using normal data. Experiments using three groups of real data were conducted to compare the anomaly detection results of the proposed method with those of manual recognition. The comparison results indicated that the proposed LSTM network achieves promising results and is viable for detecting anomalies in earthquake precursor data.",
keywords = "anomaly detection, deep learning, Earthquake precursor data, LSTM-RNN, prediction model",
author = "Yin Cai and Shyu, {Mei Ling} and Tu, {Yue Xuan} and Teng, {Yun Tian} and Hu, {Xing Xing}",
year = "2019",
month = "1",
day = "1",
doi = "10.1007/s11770-019-0774-1",
language = "English (US)",
journal = "Applied Geophysics",
issn = "1672-7975",
publisher = "Higher Education Press",

}

TY - JOUR

T1 - Anomaly detection of earthquake precursor data using long short-term memory networks

AU - Cai, Yin

AU - Shyu, Mei Ling

AU - Tu, Yue Xuan

AU - Teng, Yun Tian

AU - Hu, Xing Xing

PY - 2019/1/1

Y1 - 2019/1/1

N2 - Earthquake precursor data have been used as an important basis for earthquake prediction. In this study, a recurrent neural network (RNN) architecture with long short-term memory (LSTM) units is utilized to develop a predictive model for normal data. Furthermore, the prediction errors from the predictive models are used to indicate normal or abnormal behavior. An additional advantage of using the LSTM networks is that the earthquake precursor data can be directly fed into the network without any elaborate preprocessing as required by other approaches. Furthermore, no prior information on abnormal data is needed by these networks as they are trained only using normal data. Experiments using three groups of real data were conducted to compare the anomaly detection results of the proposed method with those of manual recognition. The comparison results indicated that the proposed LSTM network achieves promising results and is viable for detecting anomalies in earthquake precursor data.

AB - Earthquake precursor data have been used as an important basis for earthquake prediction. In this study, a recurrent neural network (RNN) architecture with long short-term memory (LSTM) units is utilized to develop a predictive model for normal data. Furthermore, the prediction errors from the predictive models are used to indicate normal or abnormal behavior. An additional advantage of using the LSTM networks is that the earthquake precursor data can be directly fed into the network without any elaborate preprocessing as required by other approaches. Furthermore, no prior information on abnormal data is needed by these networks as they are trained only using normal data. Experiments using three groups of real data were conducted to compare the anomaly detection results of the proposed method with those of manual recognition. The comparison results indicated that the proposed LSTM network achieves promising results and is viable for detecting anomalies in earthquake precursor data.

KW - anomaly detection

KW - deep learning

KW - Earthquake precursor data

KW - LSTM-RNN

KW - prediction model

UR - http://www.scopus.com/inward/record.url?scp=85075146883&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85075146883&partnerID=8YFLogxK

U2 - 10.1007/s11770-019-0774-1

DO - 10.1007/s11770-019-0774-1

M3 - Article

AN - SCOPUS:85075146883

JO - Applied Geophysics

JF - Applied Geophysics

SN - 1672-7975

ER -