Review Reports - A Novel Bézier LSTM Model: A Case Study in Corn Analysis

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

This paper introduces a new machine learning model called the Bézier LSTM model. The core focus of this work is on the model itself, rather than the specific application (corn) it's being used for. Here are some recommended improvements to enhance the paper's clarity and impact:

Consider revising the title to prominently feature the Bézier LSTM model, such as "A Novel Bézier LSTM Model: A Case Study in Corn Analysis." This emphasizes the model's novelty and its application to corn data.
Expand the literature review section to provide a more comprehensive overview of LSTM models. Discuss various LSTM types, compare their strengths and weaknesses, and highlight relevant applications in different fields. This will provide context for your proposed Bézier LSTM model.
When presenting your results, prioritize the error estimation comparison across different LSTM models (including your Bézier LSTM) as the central finding. This will directly showcase the effectiveness of your novel model.

Additional Tips for Publication:

Clearly articulate the problem the Bézier LSTM model aims to solve and how it addresses limitations of existing methods.
Provide compelling evidence demonstrating the model's performance compared to other LSTMs.
Ensure the paper is well-structured, easy to follow, and uses clear language.

By incorporating these suggestions, you can strengthen your paper's focus on the Bézier LSTM model and increase its chances of publication.

Author Response

review1：Consider revising the title to prominently feature the Bézier LSTM model, such as "A Novel Bézier LSTM Model: A Case Study in Corn Analysis." This emphasizes the model's novelty and its application to corn data.

Reply 1: Thank you for pointing this out and we agree with this comment. So we changed the title.

review2：Expand the literature review section to provide a more comprehensive overview of LSTM models. Discuss various LSTM types, compare their strengths and weaknesses, and highlight relevant applications in different fields. This will provide context for your proposed Bézier LSTM model.

Reply 2: Thank you for pointing this out and we agree with this comment. Therefore, we have added references to the types of VMD+LSTM and EMD+LSTM on page 3, line 126.

review 3：When presenting your results, prioritize the error estimation comparison across different LSTM models (including your Bézier LSTM) as the central finding. This will directly showcase the effectiveness of your novel model.

Reply 3: Thank you for pointing this out and we agree with this comment. Therefore, we added VMD+LSTM as a comparison model on page 17, line 583, to show the effectiveness of our model.

Reviewer 2 Report

Comments and Suggestions for Authors

The core of this paper is the proposal of a method to improve the predictive performance of LSTM models by using Bézier curves to remove noise from time series data. By fitting the data with Bézier curves, smoother data that better reflects trends is generated, allowing for more accurate predictions by the LSTM model.

Despite this, there is insufficient literature review in the introduction regarding the use of Bézier curves in relation to time series data.
- The introduction needs to better articulate the significance of this paper by connecting it to existing research on Bézier curves and time series data.
- A more comprehensive discussion on how Bézier curves have been previously utilized in time series analysis is necessary to highlight the novelty and relevance of this work.
There is no comparison with other noise reduction techniques such as Variational Mode Decomposition (VMD).
- At least one additional comparison with a recent noise reduction method should be included.
- This comparison is crucial in the revision process to demonstrate the effectiveness of using Bézier curves.
- Without this, the paper lacks a robust validation of its proposed method's superiority.
Despite being a key concept, the explanation of Bézier curves is insufficient.
- Visual aids and detailed explanations should be provided to help readers unfamiliar with Bézier curves understand the core concept easily.
- The paper should clearly justify the use of Bézier curves for noise reduction and trend extraction, especially through comparisons as suggested in point 2.

Author Response

review1：Despite this, there is insufficient literature review in the introduction regarding the use of Bézier curves in relation to time series data.

The introduction needs to better articulate the significance of this paper by connecting it to existing research on Bézier curves and time series data.

A more comprehensive discussion on how Bézier curves have been previously utilized in time series analysis is necessary to highlight the novelty and relevance of this work.

Reply 1: Thank you for pointing this out and we agree with this comment. Therefore, we have added relevant references on page 3, line 126.

review2：There is no comparison with other noise reduction techniques such as Variational Mode Decomposition (VMD).

At least one additional comparison with a recent noise reduction method should be included.

This comparison is crucial in the revision process to demonstrate the effectiveness of using Bézier curves.

Without this, the paper lacks a robust validation of its proposed method's superiority.

Reply 2: Thank you for pointing this out and we agree with this comment. Therefore, we added VMD+LSTM as a comparison model on page 17, line 583, to show the effectiveness of our model.

review3：Despite being a key concept, the explanation of Bézier curves is insufficient.

Visual aids and detailed explanations should be provided to help readers unfamiliar with Bézier curves understand the core concept easily.

The paper should clearly justify the use of Bézier curves for noise reduction and trend extraction, especially through comparisons as suggested in point 2.

Reply 3: Thank you for pointing this out and we agree with this comment. Therefore, we have added the bezier curve generation graph on page 4, line 167, to help the reader.

Reviewer 3 Report

Comments and Suggestions for Authors

1. Introduction

Please highlight them in your introduction in the specific paragraph.

For ORIGINAL RESEARCH PAPER

-------------------------------------

There are four (4) types of novel technical results:

1) An algorithm;

2) A system construct: such as hardware design, software system, protocol,

etc.; The main goal of your revised paper is to ensure that the next person

who designs a system like yours doesn't make the same mistakes and takes

advantage of some of your best solutions. So make sure that the hard

problems (and their solutions) are discussed and the non-obvious mistakes

(and how to avoid them) are discussed;

3) A performance evaluation: obtained through analyses, simulation or

measurements; or

4) A theory: consisting of a collection of theorems.

Your final camera ready paper should focus on:

1) Describing the results in sufficient details to establish their validity;

2) Identifying the novel aspects of the results, i.e., what new knowledge is

reported and what makes it non-obvious; and

3) Identifying the significance of the results: what improvements and impact

do they suggest.

2. Method

Accepted

3. Empirical Research

Results

To enhance the clarity and comprehensiveness of the results presentation, the following improvements could be considered:

1. Detailed Descriptions: Provide more detailed descriptions of the model training processes, including hyperparameter tuning, data preprocessing steps, and any challenges encountered during the modeling process. This additional context can help readers understand the nuances of the results.

2. Expanded Comparisons: While the comparison between models is clear, expanding this section to include a more detailed discussion on why certain models performed better or worse could provide deeper insights. Discussing the strengths and weaknesses of each model in the context of the specific dataset used would add value, and add a comparison of the data splitting process why only 80:20 is displayed where 70:30 , 90:10 to justify which experiment is better .

3. Statistical Significance: Include statistical tests to determine the significance of the differences between model performances. This could help establish whether the observed improvements with the Bézier+LSTM model are statistically significant.

4. Error Analysis: Conduct and present an error analysis to identify where and why certain models failed. Highlighting specific instances where the models struggled (e.g., sudden price spikes) can provide insights into potential areas for improvement.

5. Real-world Implications: Discuss the real-world implications of the improved predictive accuracy. For instance, how would more accurate corn futures predictions impact stakeholders such as farmers, traders, and policymakers? Providing practical examples can illustrate the broader significance of the findings.

6. Visualization Enhancements: Improve the quality of the graphs/figures. Adding annotations or highlighting key trends and turning points in the graphs can make them more informative.

By addressing these areas, the presentation of the results can be made more robust, informative, and valuable to a wider audience.

For comparison of results table 6.

1. Please improve the table structure to provide separation between each evaluation models.

2. Inconsitent text size font on title of the table.

4. Conclusion :

The paper provides a comprehensive overview of various forecasting methods but for future directions: Suggesting potential research directions or addressing limitations would add depth. For instance, exploring e.g: hybrid approaches or incorporating domain-specific knowledge.

Author Response

comments1：Introduction

Please highlight them in your introduction in the specific paragraph.

For ORIGINAL RESEARCH PAPER

-------------------------------------

There are four (4) types of novel technical results:

1) An algorithm;

2) A system construct: such as hardware design, software system, protocol,

etc.; The main goal of your revised paper is to ensure that the next person

who designs a system like yours doesn't make the same mistakes and takes

advantage of some of your best solutions. So make sure that the hard

problems (and their solutions) are discussed and the non-obvious mistakes

(and how to avoid them) are discussed;

3) A performance evaluation: obtained through analyses, simulation or

measurements; or

4) A theory: consisting of a collection of theorems.

Your final camera ready paper should focus on:

1) Describing the results in sufficient details to establish their validity;

2) Identifying the novel aspects of the results, i.e., what new knowledge is

reported and what makes it non-obvious; and

3) Identifying the significance of the results: what improvements and impact

do they suggest.

Reply 1: Thank you for pointing this out and we agree with this comment. Therefore, we have added the bezier curve generation graph on page 4, line 167, to help readers, the parameter selection of the LSTM model on page 12, line 458, and the relevant references on page 3, line 126.

comments2：Results

To enhance the clarity and comprehensiveness of the results presentation, the following improvements could be considered:

Detailed Descriptions: Provide more detailed descriptions of the model training processes, including hyperparameter tuning, data preprocessing steps, and any challenges encountered during the modeling process. This additional context can help readers understand the nuances of the results.

Expanded Comparisons: While the comparison between models is clear, expanding this section to include a more detailed discussion on why certain models performed better or worse could provide deeper insights. Discussing the strengths and weaknesses of each model in the context of the specific dataset used would add value, and add a comparison of the data splitting process why only 80:20 is displayed where 70:30 , 90:10 to justify which experiment is better .

Statistical Significance: Include statistical tests to determine the significance of the differences between model performances. This could help establish whether the observed improvements with the Bézier+LSTM model are statistically significant.

Error Analysis: Conduct and present an error analysis to identify where and why certain models failed. Highlighting specific instances where the models struggled (e.g., sudden price spikes) can provide insights into potential areas for improvement.

Real-world Implications: Discuss the real-world implications of the improved predictive accuracy. For instance, how would more accurate corn futures predictions impact stakeholders such as farmers, traders, and policymakers? Providing practical examples can illustrate the broader significance of the findings.

Visualization Enhancements: Improve the quality of the graphs/figures. Adding annotations or highlighting key trends and turning points in the graphs can make them more informative.

Reply 2: Thank you for pointing this out and we agree with this comment. Therefore, we added VMD+LSTM as a comparison model on page 17, line 583, to show the effectiveness of our model.

comments3：For comparison of results table 6.

1. Please improve the table structure to provide separation between each evaluation models.

2. Inconsitent text size font on title of the table.

Reply 3: Reply 3: Thank you for pointing this out and we agree with this comment. So we improved the structure of Table 6 on page 19, line 625, to unify the font size.

comments4. Conclusion :

Reply 4: Thank you for pointing this out and we agree with this comment. Therefore, we added the affirmation of the hybrid model of noise mitigation algorithm + prediction algorithm on page 19, line 646.

Round 2

Reviewer 2 Report

Comments and Suggestions for Authors

The authors responded to all my concerns.

Author Response

Thank you for your suggestions before, which helped us to better revise the paper.

Reviewer 3 Report

Comments and Suggestions for Authors

Need Expanded Comparisons or justify : While the comparison between models is clear, expanding this section to include a more detailed discussion on why certain models performed better or worse could provide deeper insights. Discussing and add a comparison of the data splitting process why only 80:20 is displayed where 70:30 , 90:10 to justify which experiment is better .

Author Response

comment1：Need Expanded Comparisons or justify : While the comparison between models is clear, expanding this section to include a more detailed discussion on why certain models performed better or worse could provide deeper insights. Discussing and add a comparison of the data splitting process why only 80:20 is displayed where 70:30 , 90:10 to justify which experiment is better .

reply1：Thank you for pointing this out and we agree with this comment, so we have added the reason for dividing the dataset on page 11, line 409, and the discussion of model performance on page 19, line 637.