| 郭崧.基于SVM-RF的边坡稳定性研究[J].矿产勘查,2026,17(S1):316-323 |
| 基于SVM-RF的边坡稳定性研究 |
| Research on slope stability based on SVM-RF |
| 投稿时间：2026-03-16 修订日期：2026-04-17 |
| DOI：10.20008/j.kckc.2026S1036 |
| 中文关键词: 边坡稳定性 机器学习 相关性分析 异常值处理 |
| 英文关键词: slope stability machine learning correlation analysis outlier handling |
| 基金项目:本文受中煤湖北地质勘察基础工程有限公司项目“华新骨料（武穴）有限公司一体化二期3000万吨骨料项目边坡设计”（2022-3027）、“S201浠水县黄溪冲危险路段安全整治提升工程边坡处治方案设计”（2024-3073）联合资助。 |
|
| 摘要点击次数: 4 |
| 全文下载次数: 2 |
| 中文摘要: |
| 为提升边坡稳定性预测的可靠性，本文研究了异常值处理对机器学习模型性能的优化作用，并系统比较了多种算法在相同数据条件下的预测效果。结果显示，经异常值处理后的数据集可有效提升模型训练效率、缩短收敛时间并改善预测精度。相关性分析表明，所选特征间未出现严重多重共线性，适于构建预测模型。在各算法中，SVM-RF模型的训练AUC最优，且测试结果与训练表现高度一致，说明其具有良好的泛化能力与鲁棒性。结合三峡库区典型边坡案例的工程验证进一步表明，该模型能有效拟合地质参数与边坡稳定状态间的非线性关系，具有较高的分类准确性与工程适用性。 |
| 英文摘要: |
| To improve the reliability of slope stability prediction， this study investigates the role of outlier processing in optimizing the performance of machine learning models and systematically compares the predictive performance of multiple algorithms under identical data conditions. The results indicate that the outlier-processed dataset effectively enhances model training efficiency， reduces convergence time， and improves prediction accuracy. Correlation analysis reveals no significant multicollinearity among the selected features， confirming their suitability for constructing predictive models. Among the tested algorithms， the SVM-RF model achieved the highest training AUC， with its test results demonstrating high consistency with training performance， indicating strong generalization capability and robustness. Further validation through a typical slope case study in the Three Gorges Reservoir area shows that the model effectively fits the nonlinear relationship between geological parameters and slope stability， exhibiting high classification accuracy and engineering applicability. |
|
查看全文
查看/发表评论 下载PDF阅读器 |
| 关闭 |