Stability prediction of a natural and man-made slope using various machine learning algorithms

In this paper, an attempt has been made to implement various machine learning techniques to predict the factor of safety of a natural residual soil slope and a man-made overburden mine dump slope using several physical and geometrical parameters of the respective slopes. As the stability predictions of a slope, whether natural or man-made, is very complex and time-consuming, several machine learning-based algorithms like Support Vector Regressor, Artificial Neural Network, Random Forest, Gradient Boosting and Extreme Gradient Boost were selected for modelling. The results derived from the models were compared with those achieved from numerical analysis. Moreover, various performance indices such as coefficient of determination, variance account for, root mean square error, learning rate and residual error were employed to evaluate the predictive performance of the developed models. The results indicate an excellent prediction performance and ease of interpretation of tree-based algorithms like Random Forest, Gradient Boosting and Extreme Gradient Boost than linear models like Support Vector Regressor and Neural Network-based algorithm for both the slope types. The Support Vector Regressor has the least while Extreme Gradient Boost has the highest predictive performance. Also, it was observed that the efficiency of various machine learning models to predict the factor of safety was found to be superior in the case of man-made dump slope than natural residual soil slope.