Scholarlink

<p>Slope instability refers to the phenomenon where slopes or surfaces slide, collapse, or sink due to natural or human factors. Identifying key factors and implementing protective measures are crucial for preventing slope instability. This study employs adjusted coefficient of determination (<i>Adjusted R</i><sup>2</sup>) and symmetric mean absolute percentage error (<i>sMAPE</i>) as evaluation metrics, using Bayesian nested tenfold cross-validation iteration to three machine learning models. A voting ensemble method is then applied, integrating Shapley Additive Explanations (SHAP) and non-negative matrix factorization to develop a predictive model for the slope safety factor. The primary factors influencing slope stability are investigated. Results indicate that the ensemble model outperforms the three Bayesian-optimized models, achieving an <i>Adjusted R</i><sup>2</sup> of 0.9272 and an <i>sMAPE</i> of 5.2193%. SHAP analysis reveals that slope height, slope angle, pore water pressure ratio, density, friction angle, and cohesion are significant factors affecting the slope safety factor. Considering these factors comprehensively allows for more accurate slope stability assessments and the implementation of appropriate preventive measures to enhance slope safety. This multifactor comprehensive analysis method provides a more scientific and perspective for slope stability evaluation. Furthermore, this paper develops a slope safety factor prediction platform, offering a reliable tool for predicting slope hazards, analyzing instability, and implementing protective measures in geotechnical engineering.</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 slope instability factors: An application study of a novel interpretable ensemble model

  • 期刊论文

  • 发表日期：

    2025年5月23日

查看全文
  • Shunchuan Wu,
  • Xiaolong Wang,
  • Longqiang Han,
  • Pengbin He,
  • Jingqi Cui,
  • Haokai Shen,
  • Zhuoqing Li

摘要

Slope instability refers to the phenomenon where slopes or surfaces slide, collapse, or sink due to natural or human factors. Identifying key factors and implementing protective measures are crucial for preventing slope instability. This study employs adjusted coefficient of determination (Adjusted R2) and symmetric mean absolute percentage error (sMAPE) as evaluation metrics, using Bayesian nested tenfold cross-validation iteration to three machine learning models. A voting ensemble method is then applied, integrating Shapley Additive Explanations (SHAP) and non-negative matrix factorization to develop a predictive model for the slope safety factor. The primary factors influencing slope stability are investigated. Results indicate that the ensemble model outperforms the three Bayesian-optimized models, achieving an Adjusted R2 of 0.9272 and an sMAPE of 5.2193%. SHAP analysis reveals that slope height, slope angle, pore water pressure ratio, density, friction angle, and cohesion are significant factors affecting the slope safety factor. Considering these factors comprehensively allows for more accurate slope stability assessments and the implementation of appropriate preventive measures to enhance slope safety. This multifactor comprehensive analysis method provides a more scientific and perspective for slope stability evaluation. Furthermore, this paper develops a slope safety factor prediction platform, offering a reliable tool for predicting slope hazards, analyzing instability, and implementing protective measures in geotechnical engineering.