Scholarlink

<p>This study evaluates slope stability by predicting the factor of safety (FOS) using five machine learning (ML) algorithms: multilayer perceptron (MLP), support vector machine (SVM), k-nearest neighbors (k-NN), decision tree (DT), and random forest (RF). By integrating advanced ML techniques with rigorous statistical evaluation and a comprehensive dataset, this research bridges the gap between traditional geotechnical methods and emerging computational approaches, enhancing the accuracy, reliability, and applicability of slope stability predictions. The FOS was computed based on geotechnical properties and slope geometry, including cohesion (c), internal friction angle (∅), unit weight (γ), slope height (H), and slope degree (β). The dataset, comprising 1500 slope stability instances derived from the strength reduction method (SRM), was divided into training (80%) and testing (20%) sets. The models’ accuracies were validated by the testing set; among the evaluated models, MLP achieved the best performance, with the lowest error metrics and the highest accuracy metrics: MAE = 0.03607, MSE = 0.00280, RMSE = 0.05289, RSR = 0.12763, MAPE = 0.04054, R<sup>2</sup> = 0.99011, VAF = 98.41%, and PI = 1.92129. SVM and k-NN followed closely (i.e., SVM: MAE = 0.04213, MSE = 0.00289, RMSE = 0.05373, RSR = 0.12963, MAPE = 0.05355, R<sup>2</sup> = 0.98314, VAF = 98.36%, and PI = 1.91301; k-NN: MAE = 0.03590, MSE = 0.00292, RMSE = 0.05408, RSR = 0.13049, MAPE = 0.03974, R<sup>2</sup> = 0.98292, VAF = 98.29%, and PI = 1.91176), demonstrating robust accuracy and reliability. DT and RF models exhibited less accuracy but still performed satisfactorily. Classification performance metrics, including precision, recall, F1-score, accuracy, and ROC-AUC, further highlighted MLP's superior predictive capability, with overall good performance metrics values and an AUC of 0.99418, followed by SVM (0.99369), k-NN (0.99015), DT (0.95718), and RF (0.92879). The results underscore the efficacy of ML models, particularly MLP, in improving slope stability assessments by offering accurate and efficient predictions. This research demonstrates a significant advancement in slope stability prediction by integrating ML techniques.</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

Prediction of slope stability based on five machine learning techniques approaches: a comparative study

  • 期刊论文

  • 发表日期：

    2025年3月19日

查看全文
  • Soe Hlaing Tun,
  • Changnv Zeng,
  • Farhad Jamil

摘要

This study evaluates slope stability by predicting the factor of safety (FOS) using five machine learning (ML) algorithms: multilayer perceptron (MLP), support vector machine (SVM), k-nearest neighbors (k-NN), decision tree (DT), and random forest (RF). By integrating advanced ML techniques with rigorous statistical evaluation and a comprehensive dataset, this research bridges the gap between traditional geotechnical methods and emerging computational approaches, enhancing the accuracy, reliability, and applicability of slope stability predictions. The FOS was computed based on geotechnical properties and slope geometry, including cohesion (c), internal friction angle (∅), unit weight (γ), slope height (H), and slope degree (β). The dataset, comprising 1500 slope stability instances derived from the strength reduction method (SRM), was divided into training (80%) and testing (20%) sets. The models’ accuracies were validated by the testing set; among the evaluated models, MLP achieved the best performance, with the lowest error metrics and the highest accuracy metrics: MAE = 0.03607, MSE = 0.00280, RMSE = 0.05289, RSR = 0.12763, MAPE = 0.04054, R2 = 0.99011, VAF = 98.41%, and PI = 1.92129. SVM and k-NN followed closely (i.e., SVM: MAE = 0.04213, MSE = 0.00289, RMSE = 0.05373, RSR = 0.12963, MAPE = 0.05355, R2 = 0.98314, VAF = 98.36%, and PI = 1.91301; k-NN: MAE = 0.03590, MSE = 0.00292, RMSE = 0.05408, RSR = 0.13049, MAPE = 0.03974, R2 = 0.98292, VAF = 98.29%, and PI = 1.91176), demonstrating robust accuracy and reliability. DT and RF models exhibited less accuracy but still performed satisfactorily. Classification performance metrics, including precision, recall, F1-score, accuracy, and ROC-AUC, further highlighted MLP's superior predictive capability, with overall good performance metrics values and an AUC of 0.99418, followed by SVM (0.99369), k-NN (0.99015), DT (0.95718), and RF (0.92879). The results underscore the efficacy of ML models, particularly MLP, in improving slope stability assessments by offering accurate and efficient predictions. This research demonstrates a significant advancement in slope stability prediction by integrating ML techniques.