Scholarlink

Natural disasters such as earthquakes are the most feared due to their damage and unpredictability. Due to their complexity and lack of recognizable patterns, earthquake magnitude predictions are highly unreliable. However, results using AI-based models in this area have proven to be encouraging. This research aims to use machine learning, deep learning, and stacking models to predict earthquake magnitude. Seismic data was collected from Kirkuk using the USGS (United States Geological Survey) earthquake website for the years 2022, 2023, and 2024. The results demonstrate that this approach is the most effective in terms of accuracy. The mean absolute error (MAE) and coefficient of determination (R2) were used to validate the model. The R-squared value of the model decreased from 0.9970 in the training phase to 0.9931 in the testing phase. The MAE of the model during training was 0.0350; during testing, it was 0.0579. The results support the validity and efficiency of the model.

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

Deep Stacking Model for Earthquake Prediction in Kirkuk, Iraq

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Maral Zain AL-Abdin,
  • Amel Tuama,
  • Suhail Najm Shahab

摘要

Natural disasters such as earthquakes are the most feared due to their damage and unpredictability. Due to their complexity and lack of recognizable patterns, earthquake magnitude predictions are highly unreliable. However, results using AI-based models in this area have proven to be encouraging. This research aims to use machine learning, deep learning, and stacking models to predict earthquake magnitude. Seismic data was collected from Kirkuk using the USGS (United States Geological Survey) earthquake website for the years 2022, 2023, and 2024. The results demonstrate that this approach is the most effective in terms of accuracy. The mean absolute error (MAE) and coefficient of determination (R2) were used to validate the model. The R-squared value of the model decreased from 0.9970 in the training phase to 0.9931 in the testing phase. The MAE of the model during training was 0.0350; during testing, it was 0.0579. The results support the validity and efficiency of the model.