{"title":"开发基于机器学习的地动模型,以预测土耳其的峰值地速","authors":"Fahrettin Kuran, Gülüm Tanırcan, Elham Pashaei","doi":"10.1007/s10950-024-10239-y","DOIUrl":null,"url":null,"abstract":"<div><p>This paper introduces machine learning-based Turkiye-specific ground motion models for the geometric mean horizontal component of peak ground velocity (PGV). PGV is a significant intensity metric to measure and diagnose potential earthquake damage in structures. Reliable prediction of PGV is of essential importance in precise calculations of seismic hazard. The efficiencies, reliabilities, and capabilities of various machine learning algorithms, including Random Forest, Support Vector Machine, Linear Regression, Artificial Neural Network, Gradient Boosting, and Bayesian Ridge Regression, are evaluated and compared. The most recently compiled Turkish strong motion database, which consists of over 950 earthquakes occurring from 1983 to 2023, is used for shaping the models' ability to learn and make accurate predictions. Three feature selection methods- Least Absolute Shrinkage and Selection Operator, Recursive Feature Elimination, and Pearson’s Correlation- representing embedded, wrapper, and filter approaches, respectively, are applied to determine the most suitable estimator parameters to predict PGV. Residual analyses and statistical evaluation metrics are employed to measure the performance and effectiveness of the machine learning models. Among the algorithms applied, Gradient Boosting demonstrates exceptional success in predicting PGV, particularly when utilizing all estimator parameters (features) collectively. The Gradient Boosting model exhibits superior predictive capabilities compared to existing ground motion models. It is applicable to shallow crustal strike-slip and normal faulting earthquakes with moment magnitude ranging from 3.5 to 7.8 and Joyner and Boore distance up to 200 km.</p></div>","PeriodicalId":16994,"journal":{"name":"Journal of Seismology","volume":null,"pages":null},"PeriodicalIF":1.6000,"publicationDate":"2024-09-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":"{\"title\":\"Developing machine learning-based ground motion models to predict peak ground velocity in Turkiye\",\"authors\":\"Fahrettin Kuran, Gülüm Tanırcan, Elham Pashaei\",\"doi\":\"10.1007/s10950-024-10239-y\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"<div><p>This paper introduces machine learning-based Turkiye-specific ground motion models for the geometric mean horizontal component of peak ground velocity (PGV). PGV is a significant intensity metric to measure and diagnose potential earthquake damage in structures. Reliable prediction of PGV is of essential importance in precise calculations of seismic hazard. The efficiencies, reliabilities, and capabilities of various machine learning algorithms, including Random Forest, Support Vector Machine, Linear Regression, Artificial Neural Network, Gradient Boosting, and Bayesian Ridge Regression, are evaluated and compared. The most recently compiled Turkish strong motion database, which consists of over 950 earthquakes occurring from 1983 to 2023, is used for shaping the models' ability to learn and make accurate predictions. Three feature selection methods- Least Absolute Shrinkage and Selection Operator, Recursive Feature Elimination, and Pearson’s Correlation- representing embedded, wrapper, and filter approaches, respectively, are applied to determine the most suitable estimator parameters to predict PGV. Residual analyses and statistical evaluation metrics are employed to measure the performance and effectiveness of the machine learning models. Among the algorithms applied, Gradient Boosting demonstrates exceptional success in predicting PGV, particularly when utilizing all estimator parameters (features) collectively. The Gradient Boosting model exhibits superior predictive capabilities compared to existing ground motion models. It is applicable to shallow crustal strike-slip and normal faulting earthquakes with moment magnitude ranging from 3.5 to 7.8 and Joyner and Boore distance up to 200 km.</p></div>\",\"PeriodicalId\":16994,\"journal\":{\"name\":\"Journal of Seismology\",\"volume\":null,\"pages\":null},\"PeriodicalIF\":1.6000,\"publicationDate\":\"2024-09-05\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"0\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"Journal of Seismology\",\"FirstCategoryId\":\"89\",\"ListUrlMain\":\"https://link.springer.com/article/10.1007/s10950-024-10239-y\",\"RegionNum\":4,\"RegionCategory\":\"地球科学\",\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"Q3\",\"JCRName\":\"GEOCHEMISTRY & GEOPHYSICS\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"Journal of Seismology","FirstCategoryId":"89","ListUrlMain":"https://link.springer.com/article/10.1007/s10950-024-10239-y","RegionNum":4,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q3","JCRName":"GEOCHEMISTRY & GEOPHYSICS","Score":null,"Total":0}
Developing machine learning-based ground motion models to predict peak ground velocity in Turkiye
This paper introduces machine learning-based Turkiye-specific ground motion models for the geometric mean horizontal component of peak ground velocity (PGV). PGV is a significant intensity metric to measure and diagnose potential earthquake damage in structures. Reliable prediction of PGV is of essential importance in precise calculations of seismic hazard. The efficiencies, reliabilities, and capabilities of various machine learning algorithms, including Random Forest, Support Vector Machine, Linear Regression, Artificial Neural Network, Gradient Boosting, and Bayesian Ridge Regression, are evaluated and compared. The most recently compiled Turkish strong motion database, which consists of over 950 earthquakes occurring from 1983 to 2023, is used for shaping the models' ability to learn and make accurate predictions. Three feature selection methods- Least Absolute Shrinkage and Selection Operator, Recursive Feature Elimination, and Pearson’s Correlation- representing embedded, wrapper, and filter approaches, respectively, are applied to determine the most suitable estimator parameters to predict PGV. Residual analyses and statistical evaluation metrics are employed to measure the performance and effectiveness of the machine learning models. Among the algorithms applied, Gradient Boosting demonstrates exceptional success in predicting PGV, particularly when utilizing all estimator parameters (features) collectively. The Gradient Boosting model exhibits superior predictive capabilities compared to existing ground motion models. It is applicable to shallow crustal strike-slip and normal faulting earthquakes with moment magnitude ranging from 3.5 to 7.8 and Joyner and Boore distance up to 200 km.
期刊介绍:
Journal of Seismology is an international journal specialising in all observational and theoretical aspects related to earthquake occurrence.
Research topics may cover: seismotectonics, seismicity, historical seismicity, seismic source physics, strong ground motion studies, seismic hazard or risk, engineering seismology, physics of fault systems, triggered and induced seismicity, mining seismology, volcano seismology, earthquake prediction, structural investigations ranging from local to regional and global studies with a particular focus on passive experiments.