Performance Evaluation and Engineering Verification of Machine Learning Based Prediction Models for Slope Stability
Round 1
Reviewer 1 Report
I see the potential for publication in this manuscript. There is some novelty and I think the data is interesting for readers. In what follows there are some recommendations before acceptance of this manuscript.
- There are some technical and grrametric errors in writing. for instance, the units are incorrect. kpa-> kPa.
- In Figure 6 the dimensions are not clear.
- It recommends that the authors describe the case study in more detail.
- Change the introduction and conclusions into a more cohesive approach to better explains the results and aims of this manuscript.
Author Response
Please see the attachment
Author Response File: 
Author Response.pdf
Reviewer 2 Report
The manuscript presents the original results of applying 8 methods of regression machine learning to obtain the best prediction of the slope safety factor in calculations under various loading conditions. The paper fits well within the scope of Applied Sciences Journal. The results and conclusions of the study can be useful when using machine learning in the numerical solution of engineering-geological problems of mountain slope instability. The results are of interest to a wide range of readers-specialists in the field of engineering-geological problems.
New approach and found solution are proposed for building and evaluating a model that predicts the slope safety factor. The analysis includes eight most commonly used machines machine learning regression methods, such as support vector machine (SVM), decision tree (DT), nearest neighbor algorithm (kNN), AdaBoost(ADA), random forest (RF), artificial neural network (ANN), guided clustering algorithm (Bagging) and gradient lifting iterative decision tree (GBDT). Using various algorithms, theoretical predictions of the values of 6 parameters related to the geometry and geotechnical characteristics of each slope were obtained. Among them are slope height (h), total slope angle (β) and soil bulk density (γ), cohesion parameter (c) according to the Mohr-Coulomb yield criterion, internal friction angle (φ), soil pore moisture (ru). The results indicate that ANN and RF models have been shown better performance in predicting FOS. The obtained FOS of slopes under different water content are close, while the error of predictions is related to the choice of data set and the difference in analysis methods.
Minor deficiencies are:
Line 85 the parameter of cohesion (c) have to be marked as “C”.
Line 109/110 in the Table 1 and Figure 1, Figure 2
The dimension of parameter “c” in the Table 1 should be ( kPa)
Parameters Ø, H were not commented before.
Line 225 The “Ø” and “C” symbols are not correlated with symbol «j» and «с».
Line 242 The dimension of parameter “c” in the Table8 have to be ( kPa), “Ø” and “C” symbols are not corresponded with symbols « Ø » and «с» described in Lines 85 and 86.
Author Response
Please see the attachment
Author Response File: Author Response.pdf
Reviewer 3 Report
Please see the attached file
Comments for author File: Comments.pdf
Author Response
Please see the attachment
Author Response File: Author Response.pdf
Reviewer 4 Report
Dear authors, your study site: where is it? which are the geophysical characteristics of your study area? why this area is important to study??
You are presenting a very good work without any site explanation.
You must improve this part of your work adding a Study site section adding maps, and all the characteristics of this study area.
Small edits in the attached PDF.
All the best.
Comments for author File: Comments.pdf
Author Response
Please see the attachment
Author Response File: Author Response.pdf
Round 2
Reviewer 3 Report
Assessment of the stability of engineering-geological slopes is of great importance for monitoring risks to the safety of many artificial and natural objects. Slope stability depends on a large number of factors, so predicting slope safety is a difficult task, for which it is advisable to use models based on machine learning methods. However, due to the relative novelty of such models, their improvement and engineering verification are necessary, which is taken into account in the manuscript. Thus, the study was conducted on an actual topic and was aimed at solving an important problem for practice.
This study examined the performance of eight machine learning models predicting the slope safety factor. The novelty and reliability of the modeling results obtained by the authors is confirmed by comparison with the published data of other authors.
The manuscript corresponds to the subject of the journal.
The methodology and results of the study may be of interest to a large number of specialists in the field of slope safety and stability. The manuscript contains a large amount of data that can be used to calculate slope stability.
The study makes a positive contribution to the development of slope stability models to assess their safety. According to the results of the review, the manuscript was sufficiently finalized for publication in the journal Applied Sciences.
Reviewer 4 Report
Thank you for your corrections,
All the best.
