Scholarlink

Abstract <p>Predicting earthquake magnitudes remains a major challenge in seismology, particularly in regions like Myanmar, which lies along the highly active Sagaing Fault. Traditional methods often fall short due to the complexity of seismic data, but ML offers promising alternatives. In this study, we evaluate four ensemble ML models – RF, XGBoost, GBM, and LightGBM using a dataset of 1,891 events (<i>M</i><sub><i>w</i></sub> ≥ 4.5) from Myanmar, spanning the years 1975–2025. GBM achieved the highest tolerance accuracy at a ±0.01 threshold, the lowest MAE, and a near-perfect <i>R</i><sup>2</sup> score, consistently outperforming the other models across all tested thresholds (±0.03 to ±0.1). These results highlight the critical role of feature engineering and demonstrate GBM’s effectiveness as a robust tool for modelling complex seismic behaviour. This work underscores AI’s potential to enhance earthquake preparedness in Myanmar and other tectonically similar regions.</p> Research highlights <p><UnorderedList Mark="Bullet"> <ItemContent> <p>GBM outperforms RF, XGBoost, and LightGBM&#xa0;with 99.69% tolerance accuracy (±0.01) for earthquake magnitude prediction in Myanmar.</p> </ItemContent> <ItemContent> <p>Feature engineering&#xa0;(<i>b</i>-value, time lags, mean magnitude) significantly enhances model performance.</p> </ItemContent> <ItemContent> <p>Publicly available dataset&#xa0;enables reproducibility and future research.</p> </ItemContent> <ItemContent> <p>Computationally efficient features&#xa0;allow real-time deployment on low-cost hardware (e.g., Raspberry Pi).</p> </ItemContent> <ItemContent> <p>Potential for operational forecasting&#xa0;in tectonically similar regions.</p> </ItemContent> </UnorderedList></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

Analysis of AI models for earthquake magnitude prediction in Myanmar

  • 期刊论文

  • 发表日期：

    2025年11月28日

查看全文
  • Shivani Sharma,
  • Tanvi Sharma,
  • Nitika Sharma

摘要

Abstract

Predicting earthquake magnitudes remains a major challenge in seismology, particularly in regions like Myanmar, which lies along the highly active Sagaing Fault. Traditional methods often fall short due to the complexity of seismic data, but ML offers promising alternatives. In this study, we evaluate four ensemble ML models – RF, XGBoost, GBM, and LightGBM using a dataset of 1,891 events (Mw ≥ 4.5) from Myanmar, spanning the years 1975–2025. GBM achieved the highest tolerance accuracy at a ±0.01 threshold, the lowest MAE, and a near-perfect R2 score, consistently outperforming the other models across all tested thresholds (±0.03 to ±0.1). These results highlight the critical role of feature engineering and demonstrate GBM’s effectiveness as a robust tool for modelling complex seismic behaviour. This work underscores AI’s potential to enhance earthquake preparedness in Myanmar and other tectonically similar regions.

Research highlights

GBM outperforms RF, XGBoost, and LightGBM with 99.69% tolerance accuracy (±0.01) for earthquake magnitude prediction in Myanmar.

Feature engineering (b-value, time lags, mean magnitude) significantly enhances model performance.

Publicly available dataset enables reproducibility and future research.

Computationally efficient features allow real-time deployment on low-cost hardware (e.g., Raspberry Pi).

Potential for operational forecasting in tectonically similar regions.