Seismic indicators based earthquake predictor system using Genetic Programming and AdaBoost classification

doi:10.1016/j.soildyn.2018.04.020

Soil Dynamics and Earthquake Engineering

Volume 111, August 2018, Pages 1-7

https://doi.org/10.1016/j.soildyn.2018.04.020 Get rights and content

Highlights

•
Three seismic regions are considered for earthquake prediction research.
•
Seismic indicators, representing the geophysical state of the regions are computed.
•
Genetic Programming coupled with AdaBoost is proposed to model indicators with earthquakes.
•
Prediction results are computed and compared with the available models.

Abstract

In this study an earthquake predictor system is proposed by combining seismic indicators along with Genetic Programming (GP) and AdaBoost (GP-AdaBoost) based ensemble method. Seismic indicators are computed through a novel methodology in which, the indicators are computed to obtain maximum information regarding seismic state of the region. The computed seismic indicators are used with GP-AdaBoost algorithm to develop an Earthquake Predictor system (EP-GPBoost). The setup has been arranged to provide predictions for earthquakes of magnitude 5.0 and above, fifteen days prior to the earthquake. The regions of Hindukush, Chile and Southern California are considered for experimentation. The EP-GPBoost has produced noticeable improvement in earthquake prediction due to collaboration of strong searching and boosting capabilities of GP and AdaBoost, respectively. The earthquake predictor system shows enhanced results in terms of accuracy, precision and Matthews Correlation Coefficient for the three considered regions in comparison to contemporary results.

Introduction

Earthquakes are highly feared natural catastrophic events that pose threat to human lives and cause economic damages. The early predictions of such damaging events have a potential for saving human lives and diminish financial losses. Earthquake Predictor System (EPS) aims to generate an alarm about the occurrence of earthquakes [1]. The seismic indicators based EPS aims to predict earthquakes fifteen days prior to earthquake. The seismic indicators are computed using the temporal sequence of past earthquakes, recorded in earthquake catalog. These seismic indicators are provided to computationally intelligent algorithms to generate earthquake predictions, which eventually lead to creation of EPS regarding forthcoming earthquakes. Moreover, the contemporary literature sites numerous studies, which have employed seismic indicators in collaboration with machine learning based methods, to predict earthquake occurrences, thus findings of such studies are equally significant for the development of an earthquake predictor system.

Earthquake prediction is a challenging topic [2] and endeavors have been made to predict earthquakes for over a century [3]. Earthquake prediction approaches can be categorized into three types [4]: a) mathematical and statistical methods [5], [6], b) precursor investigations [7], [8], [9], c) machine learning methodologies [10], [11], [12]. The recent encouraging results obtained in this field of research are the outcome of interdisciplinary interaction mainly, between seismology and Computational Intelligence (CI).

Machine learning based methodologies for earthquake prediction use seismic indicators in order to develop a correlation between the indicators and subsequent earthquakes. Thus, the study relies on the temporal seismic behavior of a region. The computation of seismic indicators is an effort to express the renowned principles of Gutenberg-Richter's law, seismic energy release, foreshock frequency and seismic rate changes, in numeric form. Machine learning based earthquake prediction methodologies can also be classified into two categories, depending upon the calculation approach of seismic indicators:

a)
Computation of seismic indicators after a fixed duration of time [11], [12], [13].
b)
Computation of seismic indicators after every earthquake, inclusive of the recent earthquake [10], [14], [15].

The former approach is designed to consider a fixed duration, such as 1 month, 2 weeks, so forth, for single prediction period. It does not address the issue, if multiple earthquakes strike the same region within the single prearranged prediction period. However, this issue of making multiple predictions for a single duration can be addressed through the latter approach. A seismic event occurrence may change the internal seismic state of the region. Therefore, fresh seismic indicators are essentially computed if an earthquake strikes the region during a prediction period. A new earthquake prediction is obtained based upon latest indicators, without waiting for the end of a prearranged prediction period. The latter approach is also advantageous in terms of number of feature instances to be used for developing prediction model. The greater is the number of feature instances, the better a model is trained. Since seismic indicators are computed for every recorded earthquake, therefore greater number of feature instances are available for training of model.

EPS is a field of science that has a potential for advancements. With the advent of computer based techniques, rapid progress has been observed in research and technology. CI and machine learning techniques have been used vastly for classification and, regression to obtain solution for many problems. For example, diagnosis through medical images [16], churn prediction through customer profiling [17], automatic surveillance in videos [18], differentiation between micro seismic events and quarry blasts [19], geological interpretation of structures [20] and so forth.

In this study a novel idea of ensemble classification where Genetic Programming (GP) is evolved using boosting (GP-Adaboost), has been applied for earthquake prediction (EP-GPBoost). Seismically active regions of Hindukush, Chile and Southern California are considered in this study for modelling earthquakes and seismic indicators through GP-AdaBoost. GP-AdaBoost is a unique ensemble classifier, where searching capabilities of GP and boosting of AdaBoost are combined to develop a strong classifier. The GP's evolution is supported through boosting, where multiple GP strings are evolved per class, which act as single class classifier.

In rest of the manuscript, Section 2 contains details of the related literature. Section 3 encompasses the employed methodology, including the computation of seismic indicators along with details of GP and AdaBoost based methodology. Results and discussions can be found in Section 4.

Section snippets

Related work

This section provides the overview of the research methodologies offering the use of varied seismic indicators and precursors along with various machine learning techniques. Seismic precursory analysis has been carried out to develop earthquake prediction model through detecting anomalous patterns in these signals. It is presumed that unusual variations in seismic precursors are observed during earthquake preparation process caused by tectonic movements beneath earth surface [21]. The some of

Data and methods

The regions considered for performing research on seismic indicators based EPS are Hindukush, Chile and Southern California. A large number of earthquakes have occurred in aforementioned regions which make them interesting for earthquake prediction. In seismic indicators based EPS, the required raw dataset is temporal sequence of past seismicity for the selected regions. The past seismicity is available in the form of a catalog, which is publicly available from the United States Geological

Results and discussion

In this research EPS is modelled as a binary classification task with aim to generate prediction for earthquakes of magnitude 5.0 and greater 15 days prior to an earthquake. The results of the proposed methodology are evaluated through parameters given below.

Conclusion

In this research, seismic indicators based EP-GPBoost has been proposed. A unique methodology is devised, which encompasses the maximum information of a region through the computation of available seismic indicators. These indicators are fed to a Genetic Programming (GP) and AdaBoost (GP-AdaBoost) based ensemble classification methodology. GP-AdaBoost is a unique combination of strong searching and boosting capabilities of GP and AdaBoost, respectively. The GP-AdaBoost based model has been

References (37)

Z. Jilani et al.
Monitoring and descriptive analysis of radon in relation to seismic activity of Northern Pakistan
J Environ Radioact
(2017)
M. Awais et al.
Satellite thermal IR and atmospheric radon anomalies associated with the Haripur earthquake (Oct 2010; Mw 5.2), Pakistan
Adv Space Res
(2017)
J. Reyes et al.
Neural networks to predict earthquakes in Chile
Appl Soft Comput
(2013)
H. Adeli et al.
A probabilistic neural network for earthquake magnitude prediction
Neural Netw
(2009)
A. Morales-Esteban et al.
Earthquake prediction in seismogenic areas of the Iberian Peninsula based on computational intelligence
Tectonophysics
(2013)
X. Shang et al.
Improving microseismic event and quarry blast classification using artificial neural networks based on principal component analysis
Soil Dyn Earthq Eng
(2017)
S. Pulinets et al.
Lithosphere–atmosphere–ionosphere coupling (LAIC) model – an unified concept for earthquake precursors validation
J Asian Earth Sci
(2011)
R.A. Grant et al.
Changes in animal activity prior to a major (M = 7) earthquake in the Peruvian Andes
Phys Chem Earth Parts A/B/C
(2015)
M.H. Rafiei et al.
NEEWS: a novel earthquake early warning model using neural dynamic classification and neural dynamic optimization
Soil Dyn Earthq Eng
(2017)
F. Martínez-Álvarez et al.
Determining the best set of seismicity indicators to predict earthquakes. Two case studies: Chile and the Iberian Peninsula
Knowl-Based Syst
(2013)

G. Asencio-Cortés et al.

A sensitivity study of seismicity indicators in supervised learning to improve earthquake prediction

Knowl-Based Syst

(2016)

A. Ikram et al.

Developing an expert system based on association rules and predicate logic for earthquake prediction

Knowl-Based Syst

(2015)

A. Idris et al.

Churn prediction in telecom using random forest and PSO based data balancing in combination with various feature selection strategies

Comput Electr Eng

(2012)

C.R. Allen

Responsibilities in earthquake prediction: to the seismological Society of America, delivered in Edmonton, Alberta, May 12, 1976

Bull Seismol Soc Am

(1976)

R.J. Geller et al.

Enhanced: earthquakes cannot be predicted

Science

(1997)

R.J. Geller

Earthquake prediction: a critical review

Geophys J Int

(1997)

Q. Wang et al.

Earthquake prediction based on spatio-temporal data mining: an LSTM network approach

IEEE Trans Emerg Top Comput

(2017)

A. Sil et al.

Probabilistic models for forecasting earthquakes in the northeast region of India

Bull Seismol Soc Am

(2015)

Cited by (63)

Optimal combinations of parameters for seismic response prediction of high-speed railway bridges using machine learnings
2023, Structures
This study aims to determine the optimal number and combination of input parameters from machine learning (ML) techniques, encompassing both earthquake and bridge parameters, for effectively and efficiently predicting the seismic response of high-speed railway bridges (HSRBs). 14 earthquake parameters and 37 bridge parameters were carefully selected. To mitigate multicollinearity among earthquake parameters and identify the most influential bridge parameters, correlation matrix analysis and the “Weight” method were employed, respectively. ML models were trained using data obtained from finite element analyses of 1000 bridge-ground motion pairs. The forward floating selection algorithm sequentially (FFSAS) and the backward floating selection algorithm sequentially (BFSAS) were utilized to determine the optimal combination of ML inputs for HSRBs. The analysis results demonstrate that the ML model, when utilizing the optimal input combination, yields accurate predictions of seismic response for HSRBs under various ground motion types. The predictive performance of machine learning models improves with more input parameters, but this particular model achieves stability with only six parameters. The optimization of this parameter combination is significantly influenced by factors such as the model type, evaluation criteria, and characteristics of the bridge components. FFSAS and BFSAS methods are both applicable for identifying the optimal combination of input parameters in machine learning models.
A comprehensive review of automatic programming methods
2023, Applied Soft Computing
Automatic programming (AP) is one of the most attractive branches of artificial intelligence because it provides effective solutions to problems with limited knowledge in many different application areas. AP methods can be used to determine the effects of a system’s inputs on its outputs. Although there is increasing interest in solving many problems using these methods for a variety of applications, there is a lack of reviews that address the methods. Therefore, the goal of this paper is to provide a comprehensive literature review of AP methods. At the same time, we mention the main characteristics of the methods by grouping them according to how they represent solutions. We also try to give an outlook on the future of the field by highlighting possible bottlenecks and perspectives for the benefit of the researchers involved.
EPT: A data-driven transformer model for earthquake prediction
2023, Engineering Applications of Artificial Intelligence
The leading causes of earthquakes are crustal movements, plate movements, and collisions. In recent years, many researchers in earthquake prediction have been predicting earthquakes from historical seismic data in local areas. This approach ignores the underlying internal patterns of crustal motion, plate movement, and collisions. This paper proposes a purely data-driven deep learning model called EPT. The model uses gated feature extraction blocks (GFEB) to mine potential crustal motion and plate movement patterns from global historical seismic catalog data. It uses them to aid mainshock prediction in each local, provincial region. Experiments show that this approach improves model prediction accuracy by up to 50 percent. We also use multi-headed self-attention for the first time to capture long-term dependencies within regional time series, highlighting links between focal features and compensating for the difficulty of focusing on longer-term information in long-term time series with long short-term memory networks (LSTM). In addition, we also use the gradient harmonization mechanism classification (GHMC) loss function for the first time in earthquake prediction, effectively addressing the problem of uneven data distribution across different earthquake magnitude ranges. Finally, we validated the effectiveness of the EPT model in five provincial datasets in mainland China, and the experimental results all achieved an accuracy of over 90 percent.
A comparative study of neural network methods for first break detection using seismic refraction data over a detrital iron ore deposit
2021, Ore Geology Reviews
Accurate basement delineation is crucial for mineral deposits where formation is associated with a paleo-basement topography, such as in detrital iron ore deposits. In such environments seismic surveys provide useful data to assist the mapping of the basement interface. A key stage of the process is identification of the first arrival of elastic waves at a recording station, or so-called ‘first break’. Although a variety of automated first break detection algorithms were proposed previously for use in different seismological environments, first break detection in hard rock environments remains challenging due to low amplitude arrivals and the presence of non-stationary noise within the seismic traces. This study explores the use of three types of supervised neural networks for first break detection for a seismic refraction dataset of 15,000 traces, over a detrital iron ore deposit. A fully connected neural network (NN), a Convolutional Neural Network (CNN), and a Long-Short Term Memory recurrent neural network (LSTM) are evaluated and their predicted first breaks are compared to manually picked first breaks and predictions from the widely used Coppens’ method. The results showed that the CNN and LSTM were particularly effective for the first break detection, achieving mean square errors of 0.051 ms (with a statistical bias of 2.2 ms) and 0.058 (with a statistical bias of 21.2 ms) respectively against manual interpretation, compared to –35 ms for Coppens’ method and x ms for the NN −122 ms.
Towards advancing the earthquake forecasting by machine learning of satellite data
2021, Science of the Total Environment
Citation Excerpt :
Moreover, the applications of anomaly-evaluation applied to specific earthquakes often are with lack of understanding of the underlying physics, which may cause various or even contradictory conclusions for the same earthquake (Blackett et al., 2011a, 2011b). With the rapid development of artificial intelligence and machine learning, researchers have made progress in the domain of earth sciences (Sarkar and Mishra, 2018), especially in earthquake forecasting (Asencio-Cortés et al., 2018; Asim et al., 2018a, 2018b; Bergen et al., 2019; Hulbert et al., 2018; Lubbers et al., 2018; Rafiei and Adeli, 2017; Reyes et al., 2013; Rouet-Leduc et al., 2018). At the same time, machine learning are also very effective for spatial remote sensing data handling (Du et al., 2020).
Earthquakes have become one of the leading causes of death from natural hazards in the last fifty years. Continuous efforts have been made to understand the physical characteristics of earthquakes and the interaction between the physical hazards and the environments so that appropriate warnings may be generated before earthquakes strike. However, earthquake forecasting is not trivial at all. Reliable forecastings should include the analysis and the signals indicating the coming of a significant quake. Unfortunately, these signals are rarely evident before earthquakes occur, and therefore it is challenging to detect such precursors in seismic analysis. Among the available technologies for earthquake research, remote sensing has been commonly used due to its unique features such as fast imaging and wide image-acquisition range. Nevertheless, early studies on pre-earthquake and remote-sensing anomalies are mostly oriented towards anomaly identification and analysis of a single physical parameter. Many analyses are based on singular events, which provide a lack of understanding of this complex natural phenomenon because usually, the earthquake signals are hidden in the environmental noise. The universality of such analysis still is not being demonstrated on a worldwide scale. In this paper, we investigate physical and dynamic changes of seismic data and thereby develop a novel machine learning method, namely Inverse Boosting Pruning Trees (IBPT), to issue short-term forecast based on the satellite data of 1371 earthquakes of magnitude six or above due to their impact on the environment. We have analyzed and compared our proposed framework against several states of the art machine learning methods using ten different infrared and hyperspectral measurements collected between 2006 and 2013. Our proposed method outperforms all the six selected baselines and shows a strong capability in improving the likelihood of earthquake forecasting across different earthquake databases.
Spatiotemporally explicit earthquake prediction using deep neural network
2021, Soil Dynamics and Earthquake Engineering
Citation Excerpt :
Ultimately, after training and determining the optimal hyper-parameters for the four models based on SNN, SVM, DT, and DNN using the validation score, each trained model predicted the test data that had not been fed to the models during training and validation. In Equations (4)–(8), TP, TN, FP, and FN are defined based on the confusion table as follows [102]: TP (true positive): An earthquake occurred and predicted by the model.
Due to the complexity of predicting future earthquakes, machine learning algorithms have been used by several researchers to increase the Accuracy of the forecast. However, the concentration of previous studies has chiefly been on the temporal rather than spatial parameters. Additionally, the less correlated variables were typically eliminated in the feature analysis and did not enter the model. This study introduces and investigates the effect of spatial parameters on four ML algorithms' performance for predicting the magnitude of future earthquakes in Iran as one of the most earthquake-prone countries in the world. We compared the performances of conventional methods of Support Vector Machine (SVM), Decision Tree (DT), and a Shallow Neural Network (SNN) with the contemporary Deep Neural Network (DNN) method for predicting the magnitude of the biggest upcoming earthquake in the next week. Information Gain analysis, Accuracy, Sensitivity, Positive Predictive Value, Negative Predictive Value, and Specificity measures were exploited to investigate the outcome of using a new parameter, called Fault Density, calculated using Kernel Density Estimation and Bivariate Moran's I, on the performance of the earthquake prediction, in comparison to other commonly used parameters. We discussed the behavior of the four models while dealing with different combinations of parameters and different classes of earthquake magnitudes. The results showed promising performance of the proposed parameter for the earthquakes of high magnitudes, especially using SVM and DNN models.

View all citing articles on Scopus

View full text

Seismic indicators based earthquake predictor system using Genetic Programming and AdaBoost classification

Highlights

Abstract

Introduction

Section snippets

Related work

Data and methods

Results and discussion

Conclusion

J Environ Radioact

Adv Space Res

Appl Soft Comput

Neural Netw

Tectonophysics

Soil Dyn Earthq Eng

J Asian Earth Sci

Phys Chem Earth Parts A/B/C

Soil Dyn Earthq Eng

Knowl-Based Syst

Knowl-Based Syst

Knowl-Based Syst

Comput Electr Eng

Responsibilities in earthquake prediction: to the seismological Society of America, delivered in Edmonton, Alberta, May 12, 1976

Bull Seismol Soc Am

Enhanced: earthquakes cannot be predicted

Science

Earthquake prediction: a critical review

Geophys J Int

Earthquake prediction based on spatio-temporal data mining: an LSTM network approach

IEEE Trans Emerg Top Comput

Probabilistic models for forecasting earthquakes in the northeast region of India

Bull Seismol Soc Am