Scholarlink

<p>Assessing slope stability poses a significant challenge because of the nonlinear interactions among various geotechnical parameters. Traditional analytical methods often fall short of accurately capturing these complexities. This study introduces a novel hybrid machine learning framework that combines a relevance vector machine (RVM), logistic regression (LR), and genetic programming (GP) to increase the predictive accuracy of slope stability classification and probabilistic regression. A dataset comprising 444 slope case records with six key geotechnical attributes was used to build and validate the models. GP was employed to generate probabilistic outputs that serve as new features for classification by the LR and RVM models. Compared with standalone models, the integrated structure offers improved learning capabilities and generalizability. An evaluation using multiple performance metrics demonstrated that the RVM achieved a classification accuracy of 84.56% for training data and 78.19% for testing data, whereas the GP model yielded an R<sup>2</sup> of 0.89 and an NSE of 0.87 for regression predictions. The results confirm that the hybrid framework provides a robust and interpretable solution for practical slope stability prediction tasks.</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

Modeling of slope stability for predicting classification steps and effective performance integrating supervised machine learning approaches

  • 期刊论文

  • 发表日期：

    2025年7月10日

查看全文
  • Furquan Ahmad,
  • Rahul kumar,
  • Divesh Ranjan Kumar,
  • Pijush Samui,
  • Warit Wipulanusat

摘要

Assessing slope stability poses a significant challenge because of the nonlinear interactions among various geotechnical parameters. Traditional analytical methods often fall short of accurately capturing these complexities. This study introduces a novel hybrid machine learning framework that combines a relevance vector machine (RVM), logistic regression (LR), and genetic programming (GP) to increase the predictive accuracy of slope stability classification and probabilistic regression. A dataset comprising 444 slope case records with six key geotechnical attributes was used to build and validate the models. GP was employed to generate probabilistic outputs that serve as new features for classification by the LR and RVM models. Compared with standalone models, the integrated structure offers improved learning capabilities and generalizability. An evaluation using multiple performance metrics demonstrated that the RVM achieved a classification accuracy of 84.56% for training data and 78.19% for testing data, whereas the GP model yielded an R2 of 0.89 and an NSE of 0.87 for regression predictions. The results confirm that the hybrid framework provides a robust and interpretable solution for practical slope stability prediction tasks.