Review Reports - Multivariate Air Quality Forecasting with Residual Nested LSTM Neural Network Based on DSWT

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

I think the author's literature is very interesting, but improving it can further enhance its quality. The specific modification suggestions are as follows:

1.The author described the innovation points in the introduction, but did not explicitly identify the existing research gaps. This part needs to be rephrased

2.The author needs to conduct a literature review. Specifically, it is divided into a literature review of the research content and a literature review of the research methods. This helps to clarify the research context

3. The existing "Theoretical background" content is full of model introductions, and the author needs to rename the title

4.The use of RMSE and MAE usually requires literature review. Recommended to the author https://doi.org/10.3390/systems11080392

5.The experimental results are sufficient, but lack discussion. The author needs to compare with existing literature and point out the theoretical contribution and practical significance of the research

Author Response

1.The author described the innovation points in the introduction, but did not explicitly identify the existing research gaps. This part needs to be rephrased

Thank you for pointing this out. We agree with this comment. Therefore, we have clearly pointed out in the introduction the research gaps in three aspects compared to the models in previous studies. This change can be found in lines 143 - 159 of the revised manuscript.

Thank you for pointing this out. We agree with this comment. Therefore, we have conducted a comprehensive literature review, which includes both the literature review of research content and the literature review of research methods. This approach helps to clarify the research background. This change can be found in lines 104 - 141 of the revised manuscript.

3.The existing "Theoretical background" content is full of model introductions, and the author needs to rename the title

Thank you for pointing this out. We agree with this comment. Therefore, we have changed the title of "Theoretical Background" to "Model Overview" because this one is more about explaining the structure of the model, and the title "Model Overview" is more appropriate. This change can be found in line 187 of the revised manuscript.

4.The use of RMSE and MAE usually requires literature review. Recommended to the author https://doi.org/10.3390/systems11080392

Thank you for pointing this out. We agree with this comment. Therefore, we have used the Urban Advanced Manufacturing Development Evaluation and Forecasting System as a reference to review the literature on RMSE and MAE. This change can be found in line 479 of the revised manuscript.

5.The results of the experiment were sufficient, but there was a lack of discussion. The authors need to make comparisons with the existing literature and indicate the theoretical contribution and practical significance of the research

Thank you for pointing this out. We agree with this comment. Therefore, we have analyzed the data in section 5.3, compared it with several existing baseline models, and compared it with several existing data decomposition techniques to analyze its advantages and disadvantages, and pointed out the theoretical contribution and practical significance of this study. This change can be found in lines 526-568 of the revised manuscript.

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

In this paper, the author introduces a Multivariate Air Quality Forecasting model using a Residual Nested LSTM Neural Network based on the DSWT model. The paper provides a comprehensive description of the relevant algorithms and conducts experiments using pertinent data. It also includes comparative and ablation experiments. The results demonstrate the effectiveness of the proposed algorithms.

Some comments are as follows.

1. In the fields of meteorology and environmental science, greater emphasis is typically placed on extreme scenarios. However, extreme value samples are often scarce, which can lead to issues such as data imbalance and lower prediction accuracy for these extreme cases. It is recommended to include an analysis of the results for extreme values or an assessment of performance across different levels of scenarios, to verify the algorithm's robustness under varied conditions.

2. It is advisable to have the manuscript carefully edited by a professional with expertise in technical English to enhance the language quality.

Author Response

1.In the fields of meteorology and environmental science, greater emphasis is typically placed on extreme scenarios. However, extreme value samples are often scarce, which can lead to issues such as data imbalance and lower prediction accuracy for these extreme cases. It is recommended to include an analysis of the results for extreme values or an assessment of performance across different levels of scenarios, to verify the algorithm's robustness under varied conditions.

Thank you for pointing this out. We agree with this comment. Therefore, we have separately predicted samples with PM2.5 concentrations greater than 250 μ g/m ³ as extreme values and compared them with the overall performance to verify the robustness of the algorithm under different conditions.This change can be found in lines 647 - 659 of the revised manuscript.

2.It is advisable to have the manuscript carefully edited by a professional with expertise in technical English to enhance the language quality.

Thank you for pointing this out. We agree with this comment. This change can be found in lines 0 - 20, 143 - 159, 406 - 435, 488 - 506 and 668 - 683 of the revised manuscript.

Author Response File: Author Response.pdf

Reviewer 3 Report

Comments and Suggestions for Authors

To accurately predict AQI, the manuscript proposes a DSWT-Residual NLSTM model that integrates DSWT data decomposition techniques with a Res NLSTM model. The experimental results demonstrate that the proposed method has higher accuracy and stability. It is an interesting study for readers. However, the manuscript requires several clarifications and revisions before it can be considered for publication:

1. It should be clearly stated what the past researchers did, what their limitations were, and how the current study adds new knowledge to the existing literature.

2. Define the full name of DSWT as it is not explicitly stated in the manuscript.

3. In the comparison methods, deep learning-related methods should be included rather than ablation experiments on data decomposition.

4. Highlight the best results in the experimental results table by using bold formatting.

5. For the ablation experiments, the metrics before and after ablation should be presented simultaneously.

6. The overall metrics evaluating the entire methodological process are missing.

7. In the conclusion, the limitations or potential drawbacks of the proposed technique should be discussed.

8. Condense the conclusion section to make it more concise and focused.

9. In addition, there are some minor problems with the manuscript, reflecting that the writing of this paper is not rigorous enough.

· Line 177: Correct the term "mis."

· Line 392: Add the missing "n."

Comments on the Quality of English Language

Pay attention to English expressions throughout the manuscript to ensure grammatical accuracy, clarity, and consistency.

Author Response

1.It should be clearly stated what the past researchers did, what their limitations were, and how the current study adds new knowledge to the existing literature.

Thank you for pointing this out. We agree with this comment.Therefore, we have provided additional information on what past researchers in the relevant field have done, their limitations, and what new knowledge our research has added to existing literature.This change can be found in lines 105 - 159 of the revised manuscript.

2.Define the full name of DSWT as it is not explicitly stated in the manuscript.

Thank you for pointing this out. We agree with this comment.Therefore, we have defined the full name of DSWT in the abstract.This change can be found in line 7 of the revised manuscript.

3. In the comparison methods, deep learning-related methods should be included rather than ablation experiments on data decomposition.

Thank you for pointing this out. We agree with this comment.Therefore, in section 5.3, we compared our model with deep learning models and evaluated the advantages and disadvantages of the metrics.This change can be found in lines 525 - 567 of the revised manuscript.

4.Highlight the best results in the experimental results table by using bold formatting.

Thank you for pointing this out. We agree with this comment.Therefore, we have highlighted the best results in bold format from the experimental results table in section 5.3.This change can be found in lines 524 - 525 of the revised manuscript.

5.For the ablation experiments, the metrics before and after ablation should be presented simultaneously.

Thank you for pointing this out. We agree with this comment.Therefore, after the ablation experiment, we compared the indicators before and after ablation by giving examples of SO2 indicators.This change can be found in lines 588 - 645 of the revised manuscript.

6.The overall metrics evaluating the entire methodological process are missing.

Thank you for pointing this out. We agree with this comment.Therefore, we have set the average absolute error and root mean square error of multi pollutant prediction as the overall indicators to evaluate the performance of the entire methodology process and assess the contribution of the overall model.This change can be found in lines 478 - 503 and 650 - 656 of the revised manuscript.

7.In the conclusion, the limitations or potential drawbacks of the proposed technique should be discussed.

Thank you for pointing this out. We agree with this comment.Therefore, we have discussed the limitations and potential drawbacks of the proposed technology in the conclusion. And considered the future development directions.This change can be found in lines 672 - 683 of the revised manuscript.

8.Condense the conclusion section to make it more concise and focused.

Thank you for pointing this out. We agree with this comment.Therefore, we have compressed the conclusion section and removed the previously repeated parts to make it more concise and focused.This change can be found in lines 668 - 683 of the revised manuscript.

9.In addition, there are some minor problems with the manuscript, reflecting that the writing of this paper is not rigorous enough.

Line 177: Correct the term "mis."
Thank you for pointing this out. We agree with this comment.Therefore, we have changed the incorrect mis to m is.This change can be found in line 237 of the revised manuscript.
Line 392: Add the missing "n."Thank you for pointing this out. We agree with this comment. Thank you for pointing this out. We agree with this comment.Therefore, we have added the missing n.This change can be found in line 488 of the revised manuscript.

Author Response File: Author Response.pdf

Reviewer 4 Report

Comments and Suggestions for Authors

The paper proposes a multivariate air quality (AQI) prediction method based on the discrete stationary wavelet transform (DSWT) model and residual nested long short-term memory neural network. It addresses the shortcomings of current neural network models in handling high volatility of AQI data, capturing complex nonlinear relationships, and long-term dependencies, and provides a new solution with certain innovation. The potential limitations of the proposed model were not discussed in the article. Although the model performed well in this experiment, it may be affected by factors such as data quality and real-time performance in practical applications. It is suggested that the author add an analysis of the limitations of the model in the paper and explore possible directions for future improvement to provide a more comprehensive research perspective.

Author Response

Thank you for pointing this out. We agree with this comment.Therefore, we have explained the limitations of the model in the conclusion of the paper and provided possible directions for future improvement based on these limitations.This change can be found in lines 668 - 683 of the revised manuscript.

Author Response File: Author Response.pdf

Round 2

Reviewer 4 Report

Comments and Suggestions for Authors I have reviewed the revised manuscript of the paper titled "Multivariate Air Quality Forecasting with Residual Nested LSTM Neural Network based on DSWT". The authors have made substantial efforts in addressing the concerns raised in the previous review, which is highly commendable. I am of the opinion that the paper has reached the standard for publication. I recommend that this manuscript be accepted for publication. Comments on the Quality of English Language It is clear that the authors have made a good attempt at writing in English, and the overall quality is sufficient to make the paper understandable to the international academic community. I believe that the English quality of the paper meets the basic requirements for publication.