An ensemble method based on weight voting method for improved prediction of slope stability

Abstract

This study proposes a novel ensemble method based on weighted majority voting to evaluate the stability of slope. The ensemble classifier is composed of 5 base classifiers including random forest (namely RF), logistic regression (namely LR), naive Bayes (namely NB), support vector machine (namely SVC) and backpropagation (BP). An integrated approach was developed using 213 slope cases collected from the literature and the performance of the proposed approach was validated. Selection of training parameters was carried out by the definition of safety factor (FS) and the correlation analysis of parameters. Then, a search for the optimal hyperparameters of the base classifiers is performed using a grid search algorithm combined with a five-fold crossvalidation. Weights to each base classifier is obtained by the AUC value of the training dataset. To the end, the ensemble method based on weights is established to predict the stability of slopes in this paper. It is demonstrated that the ensemble algorithm is superior to the base classifier with regard to accuracy (ACC), Kappa(K), Precision (P), recall (R), F1 score (F1), and the receiver's operating characteristic curve area under the curve (AUC). Also, the importance scores of training parameters are obtained by the random forest theory.