Prediction of the Soil Permeability Coefficient of Reservoirs Using a Deep Neural Network Based on a Dendrite Concept
Round 1
Reviewer 1 Report
I have several comments for this paper.
1. Abstract should highlight the main findings of this paper
2. Line 52-53. REFERENCE [16-20]. These citations should express one by one. The combined reference into one sentence makes readers confused. Similar to lines 69, line 73. A lot of references for one single sentence. It is not recommended for a literature review.
3. What is the research gap of this paper? The authors need in introduction
4. The main contributions need to highlight in the manuscript
5. What is the main difference between this work and existing studies?
6. The data collection needs to express in more detail. The authors can indicate the link to website that authors collected the data
7. The general workflow of this manuscript needs to add in the method section.
8. Figure 6, Figure 7, and Figure 8 need to enlarge the axis and legend for better visualization
9. Figure 10, Figure 11, and Figure 12 need to improve the resolution and description the figure
10. The reference list found some papers in Korean. It should be noticed that the reference should refer in English for readers easily to find. Please replace all references by Korean with English
For example, REFERENCE [1], [27], [30], [35], [46]
11. The total samples of data are 168 points. How can the authors determine the best training/testing ratio? In my understanding, the splitting ratio would provide a different performance. Look at Figure 7. The sample index is 49. I think this number indicated the testing performance. It is quite confused the readers how the authors determine the ratio of training and testing set. I wonder if the authors change the ratio of training and testing data. Might the accuracy of models would enhance? I suggested the authors retrain the model in different ratios of training and testing to address this point.
12. If the data set is only 168 samples. The DNN can achieve the best performance for the data. This is a good finding. I think deep learning approaches should apply to massive data samples rather than fewer samples in this proposed approach.
13. Line 201, the authors mentioned that some epoch is set to 10000. This number would take a long time for training and overfitting would happen. I think 10000 is so a large number in this case.
14. The conclusions need to revise to better express about main results as well as the key contributions to the work.
Author Response
Please see the attachment.
Author Response File: Author Response.docx
Reviewer 2 Report
This paper predicted and evaluated the soil permeability coefficient using MR, ANFIS, DNN, and DNN-T. However, the data used for prediction was not clearly decleared. Moreover, the R2 of DNN-T proposed in this work is just 0.7387 from Figure 12, which is still not very good to a scientific paper. In general, the work is not well prepared and there is a lack of clear originality.
Author Response
Comprehensive revision of the opinions of other reviewers.
Reviewer 3 Report
Manuscript ID: processes-2220908
Title: Prediction of Soil Permeability Coefficient of Reservoir using Deep Neural Network based on Dendrite Concept
Journal: processes
Comments to authors:
This study predicts and evaluates the soil permeability coefficient using a multiple regression (MR) model, an adaptive network-based fuzzy inference system (ANFIS), a general deep neural network (DNN) model, and a deep neural network using dendrite concept (DNN-T, proposed in this study). The void ratio, unit weight, and particle size were obtained from undisturbed samples of 168 collected in embankments of reservoir in south Korea as input variables for the above-mentioned models. Although extensive work has been performed but the novelty can be questioned. However, before it can be considered for publication, please address the following comments for a revision:
1) The Abstract should be enriched with the brief details of the experimental methodology. The problem to be addressed in this study should also be highlighted in the Abstract.
2) The abstract should clearly indicate the relevance of the work for international research.
3) Please highlight the novelty in the Abstract also.
4) The authors should also present some quantitative results in the Abstract.
5) English proofreading is required for grammatical mistakes and typos.
6) The last part of the introduction should conclude the limitations of the previous studies and provide the main objectives and novelties of this study. You need to clearly address the knowledge gap and provide some meaningful phrases that your study can advance the knowledge and can fill in a knowledge gap that has not been considered yet.
7) The authors are recommended to add latest relevant literature review on such works.
8) What is the need for this work? Is this work helpful for practical applications? Which applications?
9) Sentences should be added in the Abstract or Conclusion to state the implications of this study clearly.
10) What do the authors think, does a dataset of 168 samples enough to propose accurate ANN model?
11) The literature review should be improved by adding latest references and discussion.
12) Work methodologies need more discussion.
13) Results section should be defended using technical reasons and relevant references.
14) More technical discussion to the presented experimental results should be added.
15) There are no critical review/discussions before the Conclusions. Authors should add it.
16) Conclusions should be refined and briefly presented. More numerical results should be added.
17) What are the limitations of the present study? Please mention them in the manuscript.
18) The authors can add the future recommendations based on the present study.
Author Response
Please see the attachment.
Author Response File: Author Response.docx
Round 2
Reviewer 1 Report
The comments have been addressed. The paper can be published
Reviewer 2 Report
The authors have addressed satisfactorily the comments by the reviewer and the paper deserves publication in the current form.
Reviewer 3 Report
This work can be accepted in its present form.