Scholarlink

<p>Ground motion prediction equations (GMPEs) are effectively used in seismic hazard analysis to estimate peak ground accelerations (PGAs). They exhibit reduced accuracy when applied to a wide range of earthquake magnitudes and hypocentral distances, as they are often developed using datasets with limited coverage of extreme values. This study employs machine learning (ML) techniques to develop a region-specific ground motion model (GMM) for north and northeast India, capable of determining PGA accurately over a broader range of seismic source parameters. This study evaluates the performance of four widely used ML models, namely random forest (RF), extreme gradient boosting (XGB), support vector regression (SVR), and artificial neural network (ANN), to determine the most suitable approach for predicting the PGA. A dataset of 445 recorded ground motion records from 116 earthquake events is used to train and test the models. The RF model achieved the highest predictive performance with a coefficient of determination (R<sup>2</sup>) value of 0.9056. The performance indices further confirm the superiority of the RF model over the traditional GMPEs and GMMs.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Development of ground motion models using supervised learning: application to North and Northeast India

  • 期刊论文

  • 发表日期：

    2026年1月21日

查看全文
  • Ajesh Sankar,
  • Sreevalsa Kolathayar

摘要

Ground motion prediction equations (GMPEs) are effectively used in seismic hazard analysis to estimate peak ground accelerations (PGAs). They exhibit reduced accuracy when applied to a wide range of earthquake magnitudes and hypocentral distances, as they are often developed using datasets with limited coverage of extreme values. This study employs machine learning (ML) techniques to develop a region-specific ground motion model (GMM) for north and northeast India, capable of determining PGA accurately over a broader range of seismic source parameters. This study evaluates the performance of four widely used ML models, namely random forest (RF), extreme gradient boosting (XGB), support vector regression (SVR), and artificial neural network (ANN), to determine the most suitable approach for predicting the PGA. A dataset of 445 recorded ground motion records from 116 earthquake events is used to train and test the models. The RF model achieved the highest predictive performance with a coefficient of determination (R2) value of 0.9056. The performance indices further confirm the superiority of the RF model over the traditional GMPEs and GMMs.