A hybrid ensemble method for improved prediction of slope stability

New paper published on International Journal for Numerical and Analytical Methods in Geomechanics

Paper Title: A hybrid ensemble method for improved prediction of slope stability

Authors: Chongchong Qi, Xiao-lin Tang

Abstract: Accurate prediction of slope stability is a significant issue in geomechanics with many artificial intelligence (AI) techniques being utilised. However, the application of AI has not reached its full potential due to the lack of more robust algorithms. In this paper, we proposed a hybrid ensemble method for the improved prediction of slope stability using classifier ensembles and genetic algorithm (GA). Gaussian process classification, quadratic discriminant analysis, support vector machine, artificial neural networks, adaptive boosted decision trees, and k-nearest neighbours were chosen to be individual AI techniques and the weighted majority voting was used as the combination method. Validation method was chosen to be the 10-fold cross-validation and performance measures were selected to be the accuracy, the receiver operating characteristics (ROC) curve, and the area under the ROC curve (AUC). Grid search and GA were used for the hyper-parameters tuning and weights tuning respectively. The results show that the proposed hybrid ensemble method has great potential in improving the prediction of slope stability. Compared with individual classifiers, the optimum ensemble classifier (OEC) achieved the highest AUC value (0.943) and the highest accuracy (0.902) on the testing set, denoting that the predictive performance has been improved. The OEC with the Youden’s cutoff was recommended for slope stability prediction with respect to the AUC value, the accuracy, the true positive rate and the true negative rate. This research indicates the use of the classifier ensembles, rather than the search for the ideal individual classifiers, might help for the slope stability prediction.

Keywords: Slope stability; improved prediction; classifier ensembles; genetic algorithm; weighted majority voting.

Comments from reviews:

Reviewer 1: A hybrid ensemble approach based on the classifier ensemble technique and genetic algorithm is implemented by the authors to predict slope stability. The database used for training and validation of the classifiers is obtained from published case studies. There are a total of 168 slope case studies. The input variables for the classifiers include slope height, slope angle, pore water ratio, unit weight, cohesion, and internal friction angle. Output for the classifiers is the stability of a slope which is categorised into “unstable” and “stable”. An unstable slope is defined as slope with a significant soil or rock movement.  The ensemble classification framework is developed using the weighted majority voting method and the required weights are optimised by the genetic algorithm. 10-fold cross-validation is used as the validation method. Based on the comparison of several predictive performance indicators it is demonstrated that the ensemble classifier out-performs the individual classifiers in predicting slope stability. 
The paper is well written and correctly illustrated. The methodology, dataset, modelling, testing and validation processes are elaborated. The results or the proposed modelling approach are analysed and presented sufficient level of reliability and accuracy. I recommend this paper to be published subject to the following minor amendments.

Reviewer 2: The manuscript presented work on evaluation of several classifiers with parameter optimization technique for prediction of slope stability based on a collected dataset from previous publications. The six classifiers used are GP, QDA, SVM, ANN, ABDT, and k-nearest neighbors which have been presented in previously similar work. The majority voting method was adopted to specify the weight of each classifier and the GA was used for parameter optimization. The evaluation of prediction performance was evaluated by different criteria such as the OA, ROC and AUC. The weakest aspect is there is no analytical or numerical analysis in geotechnical or geomechanical point in the manuscript. Besides, some specific comments are given below.

Representative figure from this paper:

Figure 4.jpg

General architecture of weighted majority voting.

Full paper can be obtained through Email.