中国科技核心期刊
美国化学文摘社（CAS）数据库
美国EBSCO学术数据库
日本科学技术振兴机构数据库(JST)
基于SCA-GBDT的边坡稳定性预测混合模型
- 作者：
杨星雨|王宁
- 作者单位：
- （1. 紫金（长沙）工程技术有限公司；2. 中南大学资源与安全工程学院）
基金项目：
- 湖南省研究生科研创新项目（CX20230167）
详细信息：
- 作者简介：
- 杨星雨（1997—），男，助理工程师，硕士，研究方向为边坡稳定性分析；E‑mail：1749121399@qq. com
- 此文档没有上传！
Hybrid model for slope stability prediction based on SCA-GBDT
- English Author：
Yang Xingyu¹, Wang Ning²
- Unit：
- （1. Zijin (Changsha) Engineering Technology Co., Ltd.; 2. School of Resources and Safety Engineering, Central South University）
- 摘要
- 在线预览
- 参考文献
摘要:
边坡稳定性智能预测是边坡治理和边坡结构设计的重要基础。边坡稳定性评估具有复杂性和非线性，各种智能预测模型通常具有更好的性能，为研究边坡稳定性问题提供了新的方向。构建了基于正弦余弦算法（SCA）优化梯度提升树（GBDT）的边坡稳定性预测混合模型。收集436个边坡案例来建立数据库，包含6个参数（边坡高度H、边坡角B、容重y、内聚力C、内摩擦角p、孔隙水压系数r）和边坡状态数据。80%数据划分为训练集，20%数据为测试集进行测试。通过结合五重交叉验证和正弦余弦算法来调整模型的超参数。根据准确率、精确率、召回率、Fl分数、AUC 来评估所提出方法的性能。同时，研究对比了5个经典分类机器学习模型，以评估模型预测边坡稳定性的性能和适用性。结果表明：SCA能显著提高GBDT模型的性能，SCA-GBDT模型的准确率、精确率、召回率、F1分数、AUC分别为0.8860，0.877，0.915，0.896和0.955。通过SHAP算法对模型特征重要性分析，发现内摩擦角对预测结果的影响最显著。SCA-GBDT模型为预测边坡稳定性提供了一种可靠的方法，可以应用于边坡工程实践。
关键词:
边坡稳定性；预测模型；机器学习；正弦余弦算法；梯度提升树；内摩擦角
Abstract:
Intelligent prediction of slope stability is a critical foundation for slope governance and slope structure design. Due to the complexity and nonlinearity of slope stability assessment, various intelligent prediction models often exhibit superior performance, offering new directions for addressing slope stability challenges. This study constructs a slope stability prediction hybrid model based on the Sine Cosine Algorithm (SCA) optimized Gradient Boosting Decision Tree (GBDT). A database was established using 436 slope cases, comprising 6 parameters (slope height H, slope angle B, unit weight y, cohesion C, internal friction angle ç, and pore water pressure coefficient r,) along with slope stability status data. The dataset was pided into an 80 % training set and a 20 % testing set. The hyperparameters of the modelwere tuned by integrating 5-fold cross-validation with the Sine Cosine Algorithm (SCA). The performance of the proposed method was evaluated using accuracy, precision, recall rate, Fr-score, and AUC. Additionally, 5 classical classification machine learning models were compared to assess their applicability and predictive capabilities for slope stability. Results demonstrate that SCA significantly enhances the performance of the GBDT model, with the SCA-GBDT model achieving accuracy, precision, recall rate, FSm. and AUC of 0.886 0. 0.877, 0.915, 0.896, and 0.955, respectively. Characteristic importance analysis via the SHAP algorithm revealed that the internal friction angle has the most significant impact on prediction outcomes. The SCA-GBDT model provides a reliable method for slope stability prediction, applicable to practical slope engineering.
Keywords:
slope stability; prediction model; machine learning; Sine Cosine Algorithm; Gradient Boosting Decision Tree; internal friction angle
