基于ASSA-Transformer模型的边坡稳定性预测研究

Research on Slope Stability Prediction Based on the ASSA-Transformer Model

摘要: 边坡灾害具有突发性及严重危害性，所以边坡稳定性预测成为边坡工程研究的重点。传统边坡稳定性预测方法存在精度有限、泛化能力不足等问题，因此本研究提出了一种融合自适应稀疏自注意力机制(ASSA)和Transformer架构的混合智能模型(ASSA-Transformer)，提升算法模型在边坡稳定性分类任务中的预测精度。选取支持向量机(SVM)、轻量级梯度提升机(LightGBM)、Transformer、网格搜索优化Transformer (GS-Transformer)以及粒子群优化Transformer (PSO-Transformer)作为对比模型，使用准确率、精确度、召回率和F1分数作为评价指标评估模型性能，并通过混淆矩阵可视化各模型的分类结果。研究结果表明：ASSA-Transformer模型在训练集与测试集的各项评价指标上均优于其他对比模型，表现出较强的分类预测性能与泛化能力。基于置换特征重要性(PFI)算法分析得出内摩擦角、坡角及孔隙水压力为关键影响因素，本研究为边坡稳定性的预测提供了一种新方法，对边坡工程安全评估与防灾决策具有实际意义。

Abstract: Slope disasters often occur suddenly and result in severe consequences, making slope stability prediction a key focus in slope engineering research. Traditional slope stability prediction methods suffer from limitations such as limited accuracy and insufficient generalization capabilities. To enhance the predictive accuracy of algorithmic models in slope stability classification tasks, a hybrid intelligent model (ASSA-Transformer) integrating an adaptive sparse self-attention mechanism (ASSA) and a Transformer architecture is proposed. Support Vector Machine (SVM), Lightweight Gradient Boosting Machine (LightGBM), Transformer, Grid Search Optimized Transformer (GS-Transformer), and Particle Swarm Optimized Transformer (PSO-Transformer) were selected as comparison models. Model performance was evaluated using accuracy, precision, recall, and F1 score as metrics, with classification results visualized via confusion matrices. The results demonstrate that the ASSA-Transformer model outperforms all comparison models across all evaluation metrics on both the training and test datasets, exhibiting strong classification prediction performance and generalization capability. Analysis based on the Permutation Feature Importance (PFI) algorithm identifies the internal friction angle, slope angle, and pore water pressure as key influencing factors. This study provides a novel approach for predicting slope stability, holding practical significance for slope engineering safety assessment and disaster prevention decision-making.