Scholarlink

<p>In open-pit coal mining, steep slopes can make it unsafe, damage equipment, and slow down work. Slope stability relies on interconnected geometric, mechanical, and hydrological factors, which single-model predictors frequently fail to consistently capture or elucidate clearly. This research introduces a Hybrid Ensemble Slope Stability Predictor (HESSP) that amalgamates Random Forest (RF), Multi-Layer Perceptron (MLP), and Genetic Programming (GP) with majority voting to enhance robustness while preserving interpretability. We used mean imputation, Min–Max normalization, an 80:20 stratified train–test split, and 5-fold cross-validation on the training set to tune the hyperparameters. We did this on the Kaggle Slope Stability Analysis Dataset, which has 10,000 records. HESSP did better on the test set than the best standalone model (RF: 88.0% accuracy), with 92.5% accuracy, 92.1% F1-score, and 0.85 MCC. The findings suggest that the integration of diverse learners mitigates model-specific bias and yields more consistent predictions for slope stability classification. The framework establishes a pragmatic foundation for forthcoming decision-support and monitoring-focused enhancements in open-pit slope risk management.</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

Multi-factor coupling analysis of the stability of coal geological slopes in open-pit mines

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年2月4日

查看全文
  • Haodang Li,
  • Jinlong Gao,
  • Qiang Wang,
  • Huimei Zhang

摘要

In open-pit coal mining, steep slopes can make it unsafe, damage equipment, and slow down work. Slope stability relies on interconnected geometric, mechanical, and hydrological factors, which single-model predictors frequently fail to consistently capture or elucidate clearly. This research introduces a Hybrid Ensemble Slope Stability Predictor (HESSP) that amalgamates Random Forest (RF), Multi-Layer Perceptron (MLP), and Genetic Programming (GP) with majority voting to enhance robustness while preserving interpretability. We used mean imputation, Min–Max normalization, an 80:20 stratified train–test split, and 5-fold cross-validation on the training set to tune the hyperparameters. We did this on the Kaggle Slope Stability Analysis Dataset, which has 10,000 records. HESSP did better on the test set than the best standalone model (RF: 88.0% accuracy), with 92.5% accuracy, 92.1% F1-score, and 0.85 MCC. The findings suggest that the integration of diverse learners mitigates model-specific bias and yields more consistent predictions for slope stability classification. The framework establishes a pragmatic foundation for forthcoming decision-support and monitoring-focused enhancements in open-pit slope risk management.