Scholarlink

<p>Earthquakes, as a natural phenomenon, have caused significant catastrophic losses to humanity throughout history. Many machine learning and deep learning methods have been widely applied in earthquake prediction; however, the performance of these methods is often limited by redundant seismic data features and hyperparameter optimization, resulting in low prediction accuracy. To tackle these issues, we propose a Dual-Enhanced Long Short-Term Memory(LSTM) Earthquake Prediction Method Based on Improved and Hybrid Rice-Inspired Gray Wolf Optimizers(HIGWO-LSTM). Specifically, the Hybrid Rice Optimization algorithm inspired Gray Wolf Optimizer (HROGWO) algorithm is employed for feature selection of seismic data to identify the most representative subset of features. Then, the LSTM model excels at capturing long-term dependencies in time series data, thus enabling LSTM for training and predicting earthquake data. And an improved Gray Wolf Optimizer (IGWO) with Worst Individual Disturbance (WID) is used for hyperparameter optimization of the LSTM to obtain the best hyperparameter combination and improve the model’s prediction accuracy. To validate the performance of the proposed method, we use geomagnetic and seismoacoustic data from our self-developed Acoustic &amp; Electromagnetism to AI (AETA) system, and evaluate it using the Root Mean Squared Error (RMSE), Mean Absolute Error (MAE), and R-squared (R<sup>2</sup>) metrics. Experimental results show that the proposed HIGWO-LSTM earthquake magnitude prediction model outperforms state-of-the-art methods on these evaluation metrics. Due to the high-dimensional seismic data and the iterative nature of both LSTM training and metaheuristic optimization, our method demands significant computational resources. The use of dual optimizers further increases complexity, making High-Performance Computing (HPC) essential for efficient model training and real-time prediction. The reproducible code that supports the findings of this study can be accessed at <a href="https://github.com/fight123456/papercode">https://github.com/fight123456/papercode</a>.</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

A dual-enhanced long short-term memory earthquake prediction method based on improved and hybrid rice-inspired gray wolf optimizers

  • 期刊论文

  • 发表日期：

    2025年8月5日

查看全文
  • Yi Sun,
  • Ruoxuan Huang,
  • Xinchun Yi,
  • Wen Zhou,
  • Han Wang,
  • Qiyi He,
  • Zhe Ming

摘要

Earthquakes, as a natural phenomenon, have caused significant catastrophic losses to humanity throughout history. Many machine learning and deep learning methods have been widely applied in earthquake prediction; however, the performance of these methods is often limited by redundant seismic data features and hyperparameter optimization, resulting in low prediction accuracy. To tackle these issues, we propose a Dual-Enhanced Long Short-Term Memory(LSTM) Earthquake Prediction Method Based on Improved and Hybrid Rice-Inspired Gray Wolf Optimizers(HIGWO-LSTM). Specifically, the Hybrid Rice Optimization algorithm inspired Gray Wolf Optimizer (HROGWO) algorithm is employed for feature selection of seismic data to identify the most representative subset of features. Then, the LSTM model excels at capturing long-term dependencies in time series data, thus enabling LSTM for training and predicting earthquake data. And an improved Gray Wolf Optimizer (IGWO) with Worst Individual Disturbance (WID) is used for hyperparameter optimization of the LSTM to obtain the best hyperparameter combination and improve the model’s prediction accuracy. To validate the performance of the proposed method, we use geomagnetic and seismoacoustic data from our self-developed Acoustic & Electromagnetism to AI (AETA) system, and evaluate it using the Root Mean Squared Error (RMSE), Mean Absolute Error (MAE), and R-squared (R2) metrics. Experimental results show that the proposed HIGWO-LSTM earthquake magnitude prediction model outperforms state-of-the-art methods on these evaluation metrics. Due to the high-dimensional seismic data and the iterative nature of both LSTM training and metaheuristic optimization, our method demands significant computational resources. The use of dual optimizers further increases complexity, making High-Performance Computing (HPC) essential for efficient model training and real-time prediction. The reproducible code that supports the findings of this study can be accessed at https://github.com/fight123456/papercode.