The prediction of slope stability is a vital aspect of ensuring safety in geotechnical engineering, as it significantly influences the reliability of infrastructure design and disaster mitigation strategies, particularly in mountainous areas. Traditional prediction methods, such as limit equilibrium approaches and numerical modeling, are often limited by their dependence on simplified assumptions and precise geological parameters, which are notoriously challenging to measure accurately in complex field conditions. Furthermore, these deterministic methods frequently fail to account for the highly nonlinear, heterogeneous, and stochastic nature of slope failure mechanisms. Although machine learning (ML) has emerged as a promising alternative, individual ML models often exhibit limited generalization capabilities, high variance, or bias when applied to diverse geological datasets, resulting in suboptimal predictive performance. To address these ongoing challenges, this study introduces a novel, robust data-driven framework: an Attention-Enhanced Multi-Model Fusion method based on an advanced Stacking ensemble architecture. The proposed system integrates three distinct and complementary base learners—Random Forest (RF), Support Vector Machine (SVM), and XGBoost—selected for their unique capabilities in managing high-dimensional, non-linear geotechnical data. To ensure optimal performance, the hyperparameters of each base learner are meticulously tuned using Bayesian optimization, which effectively navigates the complex parameter space to mitigate the risk of overfitting. The core innovation of this research resides in the development of the meta-learner, which incorporates an advanced dual attention mechanism. In contrast to traditional stacking ensembles that employ static or linear weighting schemes, our attention-based meta-learner dynamically allocates weights to both input features and the outputs of base models. This mechanism enables the model to autonomously identify and focus on critical geotechnical indicators, such as internal friction angle and cohesion, while adaptively prioritizing the most reliable base predictor for specific geological scenarios. Comprehensive experimental evaluations underscore the superior efficacy of the proposed Attention-Stacking framework. The model achieved an impressive accuracy of 0.8604, significantly surpassing standalone base models and traditional stacking methods. A detailed performance analysis reveals a precision of 0.9206, a recall of 0.8169, and an F1-score of 0.8657, indicating an exceptional balance between minimizing false positives and accurately identifying unstable slopes. Furthermore, feature importance analysis highlights the model’s enhanced interpretability; the attention mechanism effectively emphasizes key physical parameters that align with established geotechnical theories, thereby bridging the gap between black-box AI and engineering physics. By concurrently enhancing predictive accuracy, robustness, and transparency, this study presents a reliable, adaptive solution for slope stability assessment, offering substantial potential for advancing risk management protocols and decision-making processes in contemporary geotechnical engineering practices.
为了克服单一模型在边坡稳定性预测中泛化能力不足和稳健性差等问题,集成学习已成为提升预测性能的有效途径,其通过融合多个基学习器输出,可降低模型方差和偏差,增强整体泛化能力。目前,主流集成策略包括Bagging、Boosting和Stacking这3类(Hong et al,2005;谭勇等,2024;孟凡修,2025)。其中,Stacking作为高阶集成框架,通过引入元学习器对异构基模型输出再学习,充分利用各基模型优势,提升预测精度和稳健性,在复杂工程预测中表现出良好的潜力(孔坤锋等,2025)。然而,传统的Stacking仍存在一定的局限:一是元学习器多采用静态加权机制,忽视了不同基模型在不同输入区域的局部预测能力差异;二是基模型输出缺乏置信度建模和上下文感知,可能导致元学习器过度依赖不稳定或过拟合模型;三是基学习器类型差异显著时,固定权重融合易引发信息冗余或决策冲突,限制集成系统优化潜力。
为了确保数据的真实性和可靠性,本研究采用的数据均来自现有文献(Zhou et al,2019;张伟龙,2020;Bai et al,2022;Zhu et al,2023;Huang et al,2024),文献数据涉及国内外多个地区的露天矿和水库库区等工程场景,共有643组样本,其中失稳状态数据为288组,稳定状态数据为355组。为便于更直观地理解数据集,通过绘制小提琴图来描述数据的分布情况。如图1所示,与稳定状态相比,失稳状态下的内摩擦角和黏聚力更集中于较低值。稳定边坡的重度分布整体高于不稳定边坡,根据箱型图,坡角均集中在20°~40°,不稳定状态下的孔隙水压力明显高于稳定状态下的边坡。坡高集中在0~100 m。在不同特征参数中,稳定与失稳状态下的小提琴图形态较为相似,离群值较少,表明数据集构建合理,数据质量可靠。
(2)SVM算法模型。支持向量机是一种基于统计学习理论的机器学习方法(Dietrich et al,1999),其核心思想是通过寻找一个最优超平面来实现对数据的有效划分。该算法在处理线性不可分问题时,引入核技巧,将原始数据映射到高维特征空间,使数据更容易被线性分离,本研究采用径向基函数作为核函数。同时,SVM通过最大化类别之间的间隔来提升模型的泛化能力,减少过拟合风险。图4所示为支持向量机算法示意图。
(3)XGBoost算法模型。XGBoost是一种基于梯度提升框架的集成学习算法(Mitchell et al,2017)。该算法通过构建多个弱学习器并以加法方式逐步优化模型预测结果,具有较高的预测精度和泛化能力。XGBoost的核心思想是通过迭代训练多个决策树,每棵树拟合当前模型残差,并将所有树的结果相加,从而获得最终预测值。其目标函数包括损失函数和正则化项2个部分,具体形式如下:
AzarafzaM, AkgünH, GhazifardA,et al,2021.Discontinuous rock slope stability analysis by limit equilibrium approaches—a review[J].International Journal of Digital Earth,14(12):1918-1941.
[2]
BaiG X, HouY L, WanB F,et al,2022.Performance evaluation and engineering verification of machine learning based prediction models for slope stability[J].Applied Sciences,12(15):7890.
CervantesJ, Garcia-LamontF, Rodríguez-MazahuaL,et al,2020.A comprehensive survey on support vector machine classification:applications,challenges and trends[J].Neurocomputing,408:189-215.
[5]
DietrichR, OpperM, SompolinskyH,1999.Statistical mechanics of support vector networks[J].Physical Review Letters,82(14):2975-2978.
[6]
HongW P, SongY S, LimS G,2005.Stavility evaluation of the cut slope using artificial neural network model[J].Journal of the Korean Society of Civil Engineers C,25(4C):275-283.
[7]
HuangJ X, LuD, LinW,et al,2024.Enhancing slope stability prediction through integrated PCA-SSA-SVM modeling:a case study of Longlian expressway[J].Frontiers in Earth Science,12:1429601.
[8]
KaediM, AhnC W,2017.Robust optimization using Bayesian optimization algorithm:early detection of non-robust solutions[J].Applied Soft Computing,61:1125-1138.
[9]
MaxwellA E, WarnerT A, FangF,2018.Implementation of machine-learning classification in remote sensing:an applied review[J].International Journal of Remote Sensing,39(9):2784-2817.
[10]
MitchellR, FrankE,2017.Accelerating the XGBoost algorithm using GPU computing[J].PeerJ Computer Science,3:e127.
[11]
WangY H, WangL Q, ZhangW G,et al,2024.A physics-informed machine learning solution for landslide susceptibility mapping based on three-dimensional slope stability evaluation[J].Journal of Central South University,31(11):3838-3853.
[12]
YangY K, ZhouW, JiskaniI M,et al,2023.Slope stability prediction method based on intelligent optimization and machine learning algorithms[J].Sustainability,15(2):1169.
[13]
ZhangW G, ZhangR H, WuC Z,et al,2020.State-of-the-art review of soft computing applications in underground excavations[J].Geoscience Frontiers,11(4):1095-1106.
[14]
ZhouJ, LiE M, YangS,et al,2019.Slope stability prediction for circular mode failure using gradient boosting machine approach based on an updated database of case histories[J].Safety Science,118:505-518.
[15]
ZhuW Q, ZhangS H, LiY H,et al,2023.Efficient slope reliability analysis based on representative slip surfaces:a comparative study[J].Frontiers in Earth Science,11:1100104.
DongXiaoqiang, LiuXiaofeng,et al,2024.Strict calculation method for slope stability based on limit equilibrium[J/OL].Journal of Taiyuan University of Technology,1-11[2025-11-30].
DuZhongya, LiuCui,2025.Research on SSA-BP model and its application in slope displacement prediction under data driven[J].Automation Application,66(10):168-170.
ChenYiling, ChenFeng,et al,2025.Multi-model decision level fusion of soil slope stability predictive model and interpretability analysis[J/OL].Journal of Railway Science and Engineering,1-13[2025-11-30].
LanHengxing, PengJianbing, ZhuYanbo,et al,2022.Research and prospect of geological surface processes and major disaster effects in the Yellow River Basin[J].Scientia Sinica (Terrae),52(2):199-221.
LiuLiang, XuSenmin, WangDi,et al,2024.Research on slope stability analysis and safety evaluation of open-pit mine based on optimal slope angle[J].China Mining Magazine,33():98-104.
LuLinjuan, ChenJunzhi, RenChunfang,2025.Application of Fuzzy-AHP method in the stability analysis of bedding rocky slopes[J].Gold Science and Technology,33(3):615-625.
SuZhenning, ShaoLongtan,2022.Three-dimensional slope stability based on the finite element limit equilibrium method[J].Chinese Journal of Engineering,44(12):2048-2056.
TanYong, ChenJi, YangZhongmin,et al,2024.Slope stability prediction method based on CatBoost ensemble learning[J].Science Technology and Engineering,24(30):13153-13160.
XuMengchao, WangYunmin, LiXiaoshuang,2025.Study on the impact of open-pit to underground filling mining on stope and slope stability based on MatDEM[J].Gold Science and Technology,33(6):1139-1154.
YiJianxiang, WangChenchen, ZouJinfeng,2024.Three-dimensional stability analysis approach for the rock slope based on the modified pseudo-dynamic method[J].Journal of Railway Science and Engineering,21(6):2295-2304.
ZhangMenghan, WeiJin, BianHaiding,2022.Slope stability analysis method based on machine learning:taking 618 slopes in China as examples[J].Journal of Earch Sciences and Environment,44(6):1083-1095.
[48]
张伟龙,2020.基于机器学习的边坡稳定性分析研究[D].太原:太原理工大学.Zhang Weilong,2020.The Research of slope stability analysis based on machine learning[D].Taiyuan:Taiyuan University of Technology.
TaoZhipeng, XieYutang,et al,2025.Analysis of strict upper limit variation of slope stability based on nonlinear Mohr-Coulomb failure criterion[J/OL].Journal of East China Jiaotong University,1-13[2025-11-30].
[51]
祝经中,朱飞,2024.某露天矿山高边坡稳定性分析研究[J].中国矿业,33():124-127.
[52]
ZhuJingzhong, ZhuFei,2024.Analysis and research on high slope stability of an open-pit mine[J].China Mining Magazine,33():124-127.