Integrating Self-Attention Mechanisms and ResNet for Grain Storage Ventilation Decision Making: A Study

Round 1

Reviewer 1 Report (Previous Reviewer 1)

The author attempted to address the issue raised in the previous round. However, there are the following major issues that still need to be addressed.

Major comments

1.       Line 23, ResNet().

2.       The sentence “With the rapid development of deep learning, its applications in various fields, 50

3.       such as computer vision [4], speech recognition [5], natural language processing [6], medical diagnosis [7, 8], precision agriculture [9], stock market [10] and so on…………..”

·       [8] https://www.nature.com/articles/s41598-021-03287-8

·       [9] https://www.mdpi.com/2072-4292/15/9/2450

·       [10] https://www.mdpi.com/2227-7390/8/9/1441

4.       Re-write this sentence “. One article introduced a ventilation management model for grain storage 81 based on Bayesian networks ” with proper reference.

5.       It is confusing to read from lines 85-91, are the authors talking about the existing work or their proposal?

6.       In Line 93, the sentence starts with a number [15], not look good.

7.       In Figure 2, the word “OPtimization” should be “Optimization”. Also, the flowchart is unclear as preprocessing can not include predictive modelling.

8.       Figure 5, the heading named “unnamed “ should be removed.

9.       Figure 6 is also not clear, the author can draw a table rather putting a screenshot.

10.   How many convolutions, pooling and other layers are there in CNN? Please refer to the following articles.

a.       https://www.sciencedirect.com/science/article/pii/S0010482522008642

11.   Similarly, the network architecture for LSTM and GRU must also be specified.

12.   Put the confusion matrix entry in a whole number rather than a decimal.

Model validation with train, test and validation set needs to be presented

Author Response

Dear reviewer,

Thank you for your feedback. We have made partial revisions to address your comments regarding the placement of the literature, specifically in lines 85-91, which highlight the experimental findings of previous researchers and their contribution to our work. Furthermore, we have replaced and modified the figures as per your suggestion. Regarding the introduction of the CNN, GRU, and LSTM models used in our study, we have made additions and modifications in the corresponding sections of the manuscript. Finally, we have supplemented the dataset creation section to incorporate your suggestion of including a test set, validation set, and training set.

Thank you for your valuable input, and we hope these revisions adequately address your concerns.

Author Response File: Author Response.pdf

Reviewer 2 Report (New Reviewer)

Enclosed you can find the comments applied directly to the document. In summary some of them are:

1. In the second page, the authors affirm that an extensive literature review was performed, however this work only include 22 references and in most of the cases are cited multiple references to define a concept. For instance: [4,5,6,7], [13,14], [8,9]. There is no explanation of the methods and results obtained by some of the works cited in the document so it is not clear the analysis performed to the reviewed literature.

2. There are some references bad located as the reference [15]

3. Figures 3 and 4 should be complemented because information about the position [0,0,0], dimensions and other information is not clear.

4. I suggest numbering the equations. In addition, I suggest to explain how the equations before table 2 are obtained and how these can be used. 

5. Table 2 contains multiple parameters, and some of them are not in the equation. In this sense, I suggest to explain in a better way this subsection.

6. In the model training section, the initial parameters are defined, however, an explanation about these are defined is needed and also an explanation about these can affect the performance of the algorithm. 

7. Data in the confusion matrix is not clear, please define what kind of data is in figure 12 and why the data for each cluster is so different. 

 

Comments for author File: Comments.pdf

Author Response

Dear reviewer,

Thank you for your comments. Regarding the unclear aspects of the confusion matrix, I have replaced the corresponding figure to enhance clarity. I have also made efforts to include the suggested parameters as much as possible. I greatly appreciate your advice on the citation of literature, as it is of significant importance. Thanks to your suggestions, I have identified areas for improvement in my current work, and I will strive to enhance my research capabilities and achieve better results in future endeavors. Regarding the issue with equation numbering, I apologize for any inconvenience caused by a long equation that may have affected the aesthetics of the numbering. Your understanding is greatly appreciated.

Thank you for your valuable feedback, and I hope the revisions adequately address your concerns.

Reviewer 3 Report (New Reviewer)

The topic of investigating different neural network architectures for prediction of grain ventilation conditions is interesting and of practical importance. There are however some issues that should be addressed.

1) Line 119: "Distributed fiber optic temperature measurement technology" - please explain in more detail the measurement technology, including the sensor type

2) Line 156 mentions "outlier processing". How are outliers processed?

3) The CAE abbreviation, first used in line 170, is not explained

4) Figure 9: Do the number next to v and a indicate indices (first, second and so on) or powers? If the intention was to index the values a lower index should be used.

5) In Chapter 4 Discussion, line 364 the authors mention high computational complexity of the model. Give some more details. What is the training time and model evaluation time.

6) There are results present for different model types however model parameters (size) is not given explicitly. I assume that for the ResNet_Attention case the size of the model is the same as in figure 10. What about the other models used for comparison.

7) Only one model size is used. What is the performance of the model if other sizes (smaller or larger) are used compared to other model types.

I recommend that the paper is revised before considering publication.

Author Response

Dear reviewer,

Thank you very much for your feedback. I have made the necessary revisions based on your suggestions. I have added an explanation for the abbreviation "CAE" in the manuscript. Regarding the handling of outliers, I mentioned in the preprocessing section that we remove the outliers from the data. As for the detailed implementation of the fiber optic temperature measurement technology, it is a separate research topic conducted by another researcher in our laboratory, and I am unable to provide a detailed explanation.

I have also addressed the issue of the computational complexity of the model in the paper, as you suggested. I truly appreciate your valuable suggestions regarding the model. They are extremely important to me. Thanks to your advice, I have identified areas for improvement in my current work, and I will strive to enhance my research skills and achieve more in the future.

Once again, I sincerely appreciate your valuable feedback, and I hope that my revisions meet your requirements.

Round 2

Reviewer 1 Report (Previous Reviewer 1)

All comments are addressed. Few figures are of low resolutions and some typos are there and hope to be improved during the further process.

Author Response

Comments 1: All comments are addressed. Few figures are of low resolutions and some typos are there and hope to be improved during the further process.

Response 1: Thank you very much for your suggestion. I have replaced some of the images and corrected some of the incorrect words.

Reviewer 2 Report (New Reviewer)

All my comment were reviewed and included in this new version

Author Response

Comments 1: All my comment were reviewed and included in this new version

Response 1: Thank you very much for your suggestion.

Reviewer 3 Report (New Reviewer)

The authors addressed all of my questions. The authors mentioned that the distributed fiber optic temperature measurement technology used in their work is based on another team's research. A relevant reference should be included. I have no further comments apart from this minor remark and the paper can be accepted.

Author Response

Comments 1: The authors addressed all of my questions. The authors mentioned that the distributed fiber optic temperature measurement technology used in their work is based on another team's research. A relevant reference should be included. I have no further comments apart from this minor remark and the paper can be accepted.

Response 1: Thank you very much for your feedback. Regarding the issue you mentioned about the fiber optic temperature measurement, the related paper from our team is still under review and has not been publicly released yet. Therefore, in this paper, we have cited other relevant articles from different fields to provide explanations.

This manuscript is a resubmission of an earlier submission. The following is a list of the peer review reports and author responses from that submission.

Round 1

Reviewer 1 Report

The author attempted to predict grain storage ventilation using deep learning techniques. However, the protocol is not clear and does not include sufficient details to replicate the protocol. The main concern is data preparation and model building. The detail parameter of CNN (for eg how many conv layers are used?), and the proposed attention-based network are not specified. Furthermore, the model validation and calibration are not well specified. Therefore, the manuscript is very far from being considered as a competent publication.

The paper presents a protocol without a detailed description of the dataset, model and results. Hence, the reviewer is inclined to reject the paper.

Reviewer 2 Report

Dear Authors,

The publication is interesting and quite carefully prepared, but it still needs to be refined:

- All abbreviations used in the publication should be explained

- Figure 3 should be translated into the language of the article - English

- The research method used in the work should be explained - it was not done thoroughly enough

- In my opinion, the discussion is very cursory

- However, the conclusion lacks the main conclusions, what is new in the publication?

- Where can the results of this research be applied?