Application of Machine Learning on Slope Stability Prediction: From Data to Reliable Decisions Using Interpretable Models
Application of Machine Learning on Slope Stability Prediction: From Data to Reliable Decisions Using Interpretable Models
Ghazal Hashemipour1 Mohsen Taghavijeloudar2
1) Department of Geotechnic, Faculty of Civil Engineering, Babol Noshirvani University of Technology, 47148-7313, Babol, Iran.
2) Department of Civil and Environmental Engineering, Seoul National University, 151-744, Seoul, South Korea
IOI Code
:
AACC-AAKZZG
Publication :
Third International Conference on applied researches in civil engineering, architecture and urban planning (icarcau.com/3rd)
Abstract :
The stability of embankments in geotechnical engineering is of paramount importance, as their failure can lead to catastrophic incidents, resulting in both human and financial losses, and causing adverse impacts on the economy and environment of the affected region. In this study, mashing leering (ML) was impedimenta to address a binary classification problem (stable vs. unstable) aimed at predicting embankment stability. In this context, a comprehensive study was conducted using data from 627 embankments with diverse geometric and geotechnical characteristics. To achieve this, nine ML models were developed and tested, among which the Gradient Boosting method demonstrated the best performance (Accuracy = 0.86, Precision = 0.86, Recall =0.86, F1-Score = 0.86, AUC = 0.93, CV Mean Score = 0.83, CV Std = 0.04). In addition to evaluating the performance of the optimal model, an interpretability approach was employed to ensure transparency and explainability in the model’s decision-making process, elucidating the influence of input parameters on the output. The interpretability results revealed that cohesion (C) was the most influential parameter in predicting slope stability within this ML framework. The findings of this study indicate that integrating geometric information and geotechnical properties of embankments plays a crucial role in accurately predicting their stability, thereby assisting engineers in designing safe and optimized embankments. These insights can enhance safety and durability in dam construction projects and mitigate the risk of unforeseen failures.
Keywords : slope stability geotechnics clustering machine learning