Multi-Strategy Improved Aquila Optimizer Algorithm and Its Application in Railway Freight Volume Prediction

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

This paper introduces a Multi-Strategy Improved Aquila Optimizer (MIAO) for optimizing the hyperparameters of an LSTM-based model aimed at predicting railway freight volume. MIAO is validated through comparison with multiple optimization algorithms and is applied to historical freight data. While the paper presents interesting findings, it could benefit from addressing the following points:

Key questions to consider:

1. Please explicitly state the limitations of existing research and the gaps your work is addressing. It is not entirely clear what specific gaps are being filled in the current manuscript, beyond a broad generalization.
2. The statement "some studies have shown that the use of input features with a correlation degree of 0.8 or more will improve prediction accuracy" needs references to support this claim. Additionally, do not highly correlated features essentially become redundant?
3. What are the limitations of your study?
4. In the conclusion, please mention the practical applications of your research. How can this model be applied in real-world scenarios?

Author Response

   First of all, we would like to express our sincere gratitude to the editors and anonymous reviewers for their time, efforts and recognition given to our manuscript entitled “Monthly railway freight volume forecasting based on MIAO_LSTM model”.

   Secondly, it is worth pointing out that the editors’ and reviewers’ comments and suggestions have significantly helped us improve the quality and presentation of our manuscript further. In response to their valuable and insightful feedback, we have meticulously revised the paper, addressing each comment individually, with the main revisions highlighted in Yellow in the revised manuscript and listed as follows.

Comment 1. Please explicitly state the limitations of existing research and the gaps your work is addressing. It is not entirely clear what specific gaps are being filled in the current manuscript, beyond a broad generalization.

Response: We sincerely thank the reviewer for this suggestion. The current study focuses on static datasets. The model's ability to adapt to dynamic environments, such as sudden changes in freight demand due to external factors, has not been thoroughly investigated. While meta-heuristic algorithms and deep learning models have been widely studied independently, their combined application in freight volume prediction remains underexplored. This study bridges the gap by proposing the MIAO_LSTM model, which leverages the strengths of both MIAO (optimization) and LSTM (temporal modeling) to achieve accurate and robust predictions. 

   We pointed out the limitations of existing research and expounded on the research contributions addressed in the conclusion section.

Comment 2. The statement "some studies have shown that the use of input features with a correlation degree of 0.8 or more will improve prediction accuracy" needs references to support this claim. Additionally, do not highly correlated features essentially become redundant?

Response: We sincerely thank the reviewer for this suggestion. We have collected some references, which are as follows:

[40]Wei, C., Li, H., Luo, Z., Wang, T., Yu, Y., Wu, M., ... & Yu, M. (2024). Quantitative analysis of flame luminance and explosion pressure in liquefied petroleum gas explosion and inerting: Grey relation analysis and kinetic mechanisms. Energy, 304, 132046.doi:10.1016/j.energy.2024.132046.

[41] Liu, S., Yang, Y., & Forrest, J. Y. L. (2022). Grey systems analysis: Methods, models and applications. Cham: Springer.

[42] Chang, T. C., & Lin, S. J. (1999). Grey relation analysis of carbon dioxide emissions from industrial production and energy uses in Taiwan. Journal of Environmental Management, 56(4), 247-257.

   We are considering the correlation between inputs and outputs. A high correlation indicates a large amount of valid information. On the other hand, a high correlation between inputs implies a large amount of redundant information. We have also supplemented the references.

Comment 3. What are the limitations of your study?

Response:  We sincerely thank the reviewer for this suggestion. The limitations of our study are described in detail in conclusion section.

Comment 4. In the conclusion, please mention the practical applications of your research. How can this model be applied in real-world scenarios?

Response: We sincerely thank the reviewer for this suggestion. We have made corresponding modifications to the conclusion section. In this study, the MIAO_LSTM model was applied to predict monthly railway freight volumes. The influencing factors of freight volume were used as model inputs, while the freight volume itself served as the model output. Through comparative analysis with other models, the effectiveness of the MIAO_LSTM model in railway freight volume prediction was demonstrated.

   Finally, we would like to say thanks again sincerely to anonymous reviewers and handling editor for their time and efforts spent in reviewing the manuscript.

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

Comments and sugestions in the attached file.

Comments for author File: Comments.pdf

Author Response

Dear Reviewer：

    Thank you for your review of our manuscript and for your valuable comments. In response to your suggestions, we have made the necessary revisions to the manuscript .

   Please check the attachment.We look forward to your further feedback. 

   Thank you once again for your time and efforts.

   Best regards.

   Lei Bai

Author Response File: Author Response.pdf

Reviewer 3 Report

Comments and Suggestions for Authors

1. In Line 101, when an abbreviation such as LSTM appears for the first time in the text, it is essential to provide the full form of the abbreviation instead of using it alone here.

2. In the introduction, in addition to presenting the basic model and prediction objectives, it is recommended to include information on hyperparameter optimization. This could involve discussing relevant research advancements in the literature and elaborating on the specific factors that may influence predictions when forecasting freight volume.

3. In Section 3, the formula format in the paragraph needs to be modified.

4. The text in Figure 3 is small and difficult to read, and Figure 5 lacks clarity.

5. The conclusion section appears to be short, so it is recommended to provide a more in-depth summary of the characteristics of the prediction algorithm, as well as to discuss the direction of future research, the algorithm's applicability, and related considerations.

Author Response

Dear Reviewer：    

    Thank you for your review of our manuscript and for your valuable comments. In response to your suggestions, we have made the necessary revisions to the manuscript .    

    Please check the attachment.We look forward to your further feedback.     

    Thank you once again for your time and efforts.  

    Best regards. 

    Lei Bai

Author Response File: Author Response.pdf

Reviewer 4 Report

Comments and Suggestions for Authors

By means of an enhanced Long Short-Term Memory (LSTM) neural network optimized using the Multi-strategy Improved Aquila Optimizer (MIAO), the paper "Monthly railway goods volume forecasting based on MIAO_LSTM model" solves the problem of railway goods volume forecasting. The article offers a thorough methodological approach meant to show how better the suggested model is than alternative optimization methods and conventional neural networks. From the standpoint of scientific innovation, the paper presents a sophisticated transport forecasting solution by combining LSTM with the fresh MIAO optimization technique. By better adjusting hyperparameters, the authors claim that their suggested approach beats conventional optimization methods such as PSO, GA, WOA, and SSA, hence lowering reliance on experimentally derived values. Its scientific contribution would be strengthened, nevertheless, by a more thorough debate contrasting this approach with other state-of- the-art neural network substitutes, including transformers or hybrid models. Methodological fit-wise, the paper offers a well-organized strategy that successfully solves the study issue. Extensive validation of the MIAO optimization method using several benchmark functions reveals its better performance. The authors also choose the most pertinent input features using Grey Relational Analysis (GRA), hence lowering noise and raising forecasting accuracy. Although this method is methodologically solid, the robustness of the study would be improved by a more extensive comparison with other data selection approaches, such as automated feature engineering strategies. Regarding result validity and dependability, the paper offers thorough experimental results confirming the superiority of the MIAO_LSTM model over several LSTM versions. The performance of the model is validated in the study using several accuracy measures including MAE, RMSE, MSE, MAPE, and SMAPE. The results clearly suggest that MIAO improves prediction accuracy; still, more sensitivity analysis and confidence intervals would help to test the robustness of the model under different data sets. The results could be enlarged to cover other uses of the suggested approach outside of railway goods prediction in other spheres of transportation or economic forecasting. There are some older than five-year references on railway transport predictions. Changing these references with more recent research would help to present a better picture of the present methodological developments in the subject. Deeper statistical analysis, a wider comparison with other forecasting methods, and the inclusion of the most recent research in the literature review would help the study to be even more outstanding.

A detailed list of remarks is given below:

Lines 35–36, 56–57: Insufficient justification on how the MIAO_LSTM model responds to external shocks or abrupt changes in the economy.

Lines 35–36, 58–59: There is no study of restrictions in model scalability for major railway networks.

Line 116–119: Not compared with other contemporary deep learning techniques (such as hybrid models or transformers).

Lines 125–127, 274–275: Little conversation on MIAO's computational complexity and efficiency relative to other optimizers.

Lines 129-134, 384–386: No comparison of Grey Relational Analysis (GRA) with other feature selection methods (e.g., PCA, mutual information).

Lines 135–140: Not enough reason for using LSTM above alternative time-series models (e.g., CNN-LSTM).

Lines 275–277, 469–471: Performance results lack confidence intervals or error bars.

Lines 291–295: Insufficient ablation studies to assess specific contributions of several MIAO components.

Lines 342–344, 490–491: There is no debate on whether the model keeps performance on long-term projection—that is, beyond twelve months.

Lines 343–344: No study of overfitting considers a really sophisticated model.

Lines 356–358: Not obvious if the data underwent seasonality adjustments—that is, split into trend/seasonal/residual components.

Lines 368–369: Not mentioned hyperparameter tuning techniques outside MIAO (e.g., grid search, bayesian optimization).

Lines 375–380: No validation on outside datasets outside of Chinese railway data, hence lowering generalizability.

Lines 377–380: No robustness test against varying missing data or dataset sizes.

Lines 403–407: Not obvious why one used a 0.8 threshold for feature selection.

Lines 429–431: Uncertain effect of hyperparameter adjustment on computational economy.

Lines 440–402: No justification for why some optimizers—such as SOA, GA—perform worse than MIAO.

Lines 472–474: Not one statistical test—such as Wilcoxon signed-rank test—to validate notable variations between models.

Lines 490–498: Little conversation on practicality outside of railway goods planning.

Lines 491–497: Conclusions overlook more general consequences for freight planning in favor of their excessively limited emphasis on model correctness.

Lines 498–499: Not one recommendation for future enhancements—such as hybrid models or real-time adaptive learning.

Lines 541–550: Some sources, such as those on railway forecasting, are older than five years and lack the most recent ideas.

Author Response

Dear Reviewer：    

    Thank you for your review of our manuscript and for your valuable comments. In response to your suggestions, we have made the necessary revisions to the manuscript .    

    Please check the attachment.We look forward to your further feedback.     

    Thank you once again for your time and efforts.  

    Best regards. 

    Lei Bai

Author Response File: Author Response.pdf

Round 2

Reviewer 2 Report

Comments and Suggestions for Authors

Comments in. the attached file.

Comments for author File: Comments.docx

Author Response

   First of all, we would like to express our sincere gratitude to the editors and anonymous reviewers for their time, efforts and recognition given to our manuscript entitled “Multi-Strategy Improved Aquila Optimizer Algorithm and Its Application in Railway Freight Volume Prediction”.

    Secondly, it is worth pointing out that the editors’ and reviewers’ comments and suggestions have significantly helped us improve the quality and presentation of our manuscript further. In response to their valuable and insightful feedback, we have meticulously revised the paper, addressing each comment individually, with the main revisions highlighted in Red in the revised manuscript and listed as follows.

Comment 1. Abstract: It appears somewhat lengthy; consider condensing it for clarity.

Response: We sincerely thank the reviewer for this suggestion. The abstract has been concisely refined in accordance with the requirements.

Comment 2. References (Lines 61, 63, 66, 69, 95): Ensure proper formatting and consistency throughout the paper.

Response: We sincerely thank the reviewer for this suggestion. We have carefully revised all references to ensure consistent formatting throughout the manuscript.

Comment 3. Function Formatting (Lines 211, 220): The “tanh” and other function should not be italicized in equations.

Response:  We sincerely thank the reviewer for this suggestion. We have revised the formatting of the equations.

Comment 4. Spacing (Line 392): Add a space after the period at the end of the sentence; check for similar spacing issues throughout the document.

Response: We sincerely thank the reviewer for this suggestion. We have checked the entire document and made the necessary corrections.

Comment 5. In-line Math Formatting (Lines 409–410): In-line mathematical expressions should match formula formatting. Variables must be italicized, and spacing around the math operators, e.g. “=”, should be consistent throughout the text.

Response: We sincerely thank the reviewer for this suggestion. We have checked the entire document and made the necessary corrections.

Comment 6. Reference to Table (Lines 493–494): Keep the table number and word “Table” together on the same line.

Response: We sincerely thank the reviewer for this suggestion.We have revised the manuscript as requested.

Comment 7. Caption Spacing (Table 7): Insert a space after the period in the caption.

Response: We sincerely thank the reviewer for this suggestion.We have revised the manuscript as requested.

Comment 8. Table 9 Pagination (Line 552): Ensure that the table header and at least one data row remain together before continuing the next page.

Response: We sincerely thank the reviewer for this suggestion.We have revised the manuscript as requested.

Comment 9. Table 11 Caption (Line 579): The caption must appear on the same page as the table.

Response: We sincerely thank the reviewer for this suggestion.We have revised the manuscript as requested.

Author Response File: Author Response.pdf

Reviewer 4 Report

Comments and Suggestions for Authors

Dear authors, many thanks for the answers. You answered many of my questions and remarks, but I still recommend adding these limitations (see your answers to my comments) to your paper. The reader should know why you are not using one or the other method. I hope a few sentences on my comments 3, 5, 6, 9, 12, and 13 would provide more clarity for the reader as a limitation of the research.

Author Response

   First of all, we would like to express our sincere gratitude to the editors and anonymous reviewers for their time, efforts and recognition given to our manuscript.

   It is worth pointing out that the editors’ and reviewers’ comments and suggestions have significantly helped us improve the quality and presentation of our manuscript further. In response to their valuable and insightful feedback, we have meticulously revised the paper, addressing each comment individually, with the main revisions highlighted in Red in the revised manuscript .

Author Response File: Author Response.pdf