Scholarlink

<p>Predicting seismic hazards, particularly rockbursts, in underground mining is a critical challenge for ensuring personnel safety and operational continuity. Traditional methods often fall short in providing timely and accurate warnings. This study investigates the potential of advanced machine learning techniques to address this issue by explicitly focusing on mine-scale bump hazard prediction within mining environments. The power of two robust gradient boosting algorithms, namely the XGBoost Classifier (XGBC) and the CatBoost Classifier (CATC), is leveraged as the foundational models. To further enhance their predictive capabilities and optimize their performance, these models are integrated with the Dandelion Optimizer (DO), a novel metaheuristic algorithm. This strategic integration results in two new hybrid frameworks, namely XGDO (XGBC+DO) and CADO (CATC+DO). Upon evaluation using the well-established Seismic Bumps dataset, strong performance was observed. The CADO framework demonstrated superior results, achieving an accuracy of 0.976 and a recall of 0.976, indicating high reliability in identifying hazardous events. The XGDO framework also exhibited strong predictive capability, with an accuracy of 0.960. These results are specific to the Seismic Bumps dataset, and the models are proposed strictly for mine-scale bump hazard prediction, serving as a computationally efficient tool for seismoacoustic hazard classification in this specific context. This work presents a promising and practical approach for enhancing safety monitoring systems in mining operations.</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

Extreme gradient boosting and CAT boost classification with dandelion optimization: a novel approach to seismic activity prediction

  • 期刊论文

  • 发表日期：

    2026年7月7日

查看全文
  • Boyu Zhao,
  • Yunpeng Bai,
  • Xueyan Zhao,
  • Jingxian Meng,
  • Xiuxia Yue

摘要

Predicting seismic hazards, particularly rockbursts, in underground mining is a critical challenge for ensuring personnel safety and operational continuity. Traditional methods often fall short in providing timely and accurate warnings. This study investigates the potential of advanced machine learning techniques to address this issue by explicitly focusing on mine-scale bump hazard prediction within mining environments. The power of two robust gradient boosting algorithms, namely the XGBoost Classifier (XGBC) and the CatBoost Classifier (CATC), is leveraged as the foundational models. To further enhance their predictive capabilities and optimize their performance, these models are integrated with the Dandelion Optimizer (DO), a novel metaheuristic algorithm. This strategic integration results in two new hybrid frameworks, namely XGDO (XGBC+DO) and CADO (CATC+DO). Upon evaluation using the well-established Seismic Bumps dataset, strong performance was observed. The CADO framework demonstrated superior results, achieving an accuracy of 0.976 and a recall of 0.976, indicating high reliability in identifying hazardous events. The XGDO framework also exhibited strong predictive capability, with an accuracy of 0.960. These results are specific to the Seismic Bumps dataset, and the models are proposed strictly for mine-scale bump hazard prediction, serving as a computationally efficient tool for seismoacoustic hazard classification in this specific context. This work presents a promising and practical approach for enhancing safety monitoring systems in mining operations.