Scholarlink

<p>This study presents a method for predicting mountaineering accident risks during the planning stage, using a deep learning approach with contextual data. The method integrates Temporal, Environmental, and Demographic information, converting these into textual input for BERT’s contextual understanding. The model classifies accident risks into four categories: Falls from Height, Ground-level Falls, Fatigue, and Disorientation. The data, sourced from Nagano Prefectural Police Headquarters, covers mountain accidents in Nagano Prefecture, Japan, from 2014 to 2023. Nagano, with the highest mountain accident rate in Japan, recorded 2,596 incidents. Falls from Height accounted for 34.1% (884 cases), followed by Ground-level Falls (23.6%, 613 cases), Fatigue (22.3%, 579 cases), and Disorientation (20.0%, 520 cases). BERT achieved the highest accuracy, precision, and recall by incorporating detailed temporal and environmental data. SHAP analysis highlighted key features influencing accident risks. For example, “morning” and “hotaka” were strong predictors for Falls from Height, while “noon” and “Yatsugatake range” influenced Ground-level Falls. Fatigue was linked to late afternoon activity and elderly hikers, and Disorientation to snowy, foggy conditions, and solo hiking. This study offers a robust framework for predicting mountain accident risks, aiding in improved safety measures and planning.</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

From planning to prevention: predicting mountain accident risks using pre-climb information

  • 期刊论文

  • 发表日期：

    2025年6月16日

查看全文
  • Taeko Sato,
  • Yusuke Fukazawa

摘要

This study presents a method for predicting mountaineering accident risks during the planning stage, using a deep learning approach with contextual data. The method integrates Temporal, Environmental, and Demographic information, converting these into textual input for BERT’s contextual understanding. The model classifies accident risks into four categories: Falls from Height, Ground-level Falls, Fatigue, and Disorientation. The data, sourced from Nagano Prefectural Police Headquarters, covers mountain accidents in Nagano Prefecture, Japan, from 2014 to 2023. Nagano, with the highest mountain accident rate in Japan, recorded 2,596 incidents. Falls from Height accounted for 34.1% (884 cases), followed by Ground-level Falls (23.6%, 613 cases), Fatigue (22.3%, 579 cases), and Disorientation (20.0%, 520 cases). BERT achieved the highest accuracy, precision, and recall by incorporating detailed temporal and environmental data. SHAP analysis highlighted key features influencing accident risks. For example, “morning” and “hotaka” were strong predictors for Falls from Height, while “noon” and “Yatsugatake range” influenced Ground-level Falls. Fatigue was linked to late afternoon activity and elderly hikers, and Disorientation to snowy, foggy conditions, and solo hiking. This study offers a robust framework for predicting mountain accident risks, aiding in improved safety measures and planning.