All reviews of published articles are made public. This includes manuscript files, peer review comments, author rebuttals and revised materials. Note: This was optional for articles submitted before 13 February 2023.
Peer reviewers are encouraged (but not required) to provide their names to the authors when submitting their peer review. If they agree to provide their name, then their personal profile page will reflect a public acknowledgment that they performed a review (even if the article is rejected). If the article is accepted, then reviewers who provided their name will be associated with the article itself.
Thanks to the authors for their efforts to improve the work. This version successfully addressed the concerns of the reviewers and satisfied them. It can accepted currently. Congrats!
Thanks for revising your manuscript and addressing the comments raised by the referees. As you can see, there are just some minor issues that need to be taken into account at this stage. Please read through them and provide a detailed response letter.
I have no further comments.
I have no further comments.
I have no further comments.
I have no further comments.
Clarity and Language Precision:
Issue: The manuscript contains sections where the language remains somewhat unclear or technically dense, which could hinder comprehension for readers not deeply familiar with the specific area of neural networks.
Suggestion: It would be beneficial to simplify the language in the more complex sections and avoid jargon where possible. Additionally, consider a thorough review by a native English speaker who specializes in technical writing to ensure clarity and precision in language.
Integration and Contextualization of Literature:
Issue: While the literature review is comprehensive, the integration of new references occasionally lacks a clear connection to the specific advancements made by ProsperNN.
Suggestion: Enhance the discussion on how ProsperNN addresses the gaps identified in the literature more explicitly. A side-by-side comparison or a dedicated section discussing the evolution of models leading up to ProsperNN could provide clearer context and highlight the manuscript's contributions more distinctly.
Graphical Representations and Data Presentation:
Issue: Some figures and tables, although informative, suffer from low resolution or lack comprehensive legends, making them less effective.
Suggestion: Improve the resolution and detail of all visual aids. Ensure that each figure and table includes a detailed legend explaining all elements and abbreviations used. This will not only aid in understanding but also enhance the professional presentation of the manuscript.
Accessibility and Transparency of Data:
Issue: The manuscript mentions data availability but does not provide straightforward access links or detailed instructions on how to obtain the data, which is critical for reproducibility and transparency.
Suggestion: Provide direct access links to the raw data and any supplementary material. If data cannot be shared openly due to privacy or proprietary restrictions, please include a clear explanation and possibly an anonymized version of the data.
Technical Details and Definitions:
Issue: Some technical terms and model components are introduced without sufficient definition or explanation, which could alienate readers who are not experts in the field.
Suggestion: Include a glossary of terms or a dedicated section that defines all technical terms and model components at their first occurrence in the text. This will make the manuscript more accessible to a broader audience and uphold the professional standards of communication.
Original primary research within Aims and Scope of the journal:
Issue: While the research fits within the aims and scope of the journal, the manuscript could more explicitly articulate how the ProsperNN package addresses a specific knowledge gap in the field of neural network applications for time series forecasting.
Suggestion: To strengthen the manuscript, please clarify the unique aspects of ProsperNN compared to existing frameworks in the introduction and summary. It would be beneficial to detail the specific problems these models solve and how they improve upon or differ from existing solutions.
Research question well defined, relevant & meaningful. It is stated how research fills an identified knowledge gap:
Issue: The research question is relevant; however, the manuscript could better specify how the findings contribute to the broader scientific knowledge beyond technical improvements. The implications for practical applications in real-world scenarios could be more thoroughly explored.
Suggestion: Expand the discussion on the implications of your research findings for the field, particularly how they can be applied in practice. Providing case studies or examples of real-world applications of ProsperNN could significantly enhance the relevance and impact of your work.
Rigorous investigation performed to a high technical & ethical standard:
Issue: The technical rigor of the investigation is evident; however, there is scant mention of the ethical considerations relevant to the research. While the study may not directly involve sensitive data, discussing the ethical dimensions related to data usage, especially in predictive modeling, would strengthen the manuscript.
Suggestion: Include a section on ethical considerations, particularly focusing on data privacy, model bias, and the potential implications of predictive inaccuracies. This discussion could also cover any steps taken to mitigate risks associated with the deployment of these neural network models.
Methods described with sufficient detail & information to replicate:
Issue: The methods section is detailed but lacks some specifics that would enable replication of the research by other investigators. For instance, the exact parameter settings and the software versions used are not thoroughly documented.
Suggestion: Provide detailed descriptions of all experimental setups, including hyperparameters, software versions, and hardware specifications used in the experiments. If possible, include a supplementary material section or an online repository link where researchers can access the complete code and data set configurations.
Impact and novelty not assessed. Meaningful replication encouraged where rationale & benefit to literature is clearly stated:
Issue: The manuscript does not adequately assess the impact and novelty of ProsperNN beyond its technical advancements. While the replication of results using the package is mentioned, the rationale and broader benefits of these replications to the academic and practical fields are not clearly articulated.
Suggestion: Emphasize the novel aspects of ProsperNN by detailing how its features address specific limitations of existing models. Discuss the potential broader impacts of your findings on the field, such as improvements in efficiency, accuracy, or resource usage. Clearly state the value of replication studies by explaining how they validate the findings and extend the usability of ProsperNN in various scenarios.
All underlying data have been provided; they are robust, statistically sound, & controlled:
Issue: While the manuscript claims that the data supporting the conclusions are available, it lacks detailed information on how to access this data. Additionally, there is limited discussion on the robustness and statistical soundness of the data used.
Suggestion: Ensure that all data used in your study is uploaded to a discipline-specific repository and include links or DOIs in the manuscript. Enhance the data presentation by including a detailed analysis of the data robustness, possibly through supplementary statistical tests or validation methods. Describe any control mechanisms used to ensure the integrity of the data.
Conclusions are well stated, linked to original research question & limited to supporting results:
Issue: The conclusions drawn in the manuscript occasionally extend beyond what is directly supported by the results. Specifically, some claims about the predictive capabilities of ProsperNN may imply a causative understanding which is not supported by the data presented.
Suggestion: Tighten the conclusions to reflect only what is directly supported by the experimental results. If causal claims are necessary, ensure they are backed by appropriate experimental designs that can isolate and test these causal relationships. Clearly differentiate between correlation and causation in the text, providing a more rigorous justification for any causal claims made.
All comments have been added in detail to the last section.
All comments have been added in detail to the last section.
All comments have been added in detail to the last section.
Review Report for PeerJ Computer Science
(Introducing ProsperNN - A Python package for forecasting with neural networks)
Both the changes made to the paper and the responses given to the reviewer comments are sufficient and appropriate. For these reasons, I recommend that the paper be accepted.
Thanks for submitting your work to PeerJ Computer Science. Five referees read it and found some merits. However, they all raised some concerns about the experimental design and validity of the findings. Moreover, you need to make sure that your recommended package ensures the required readability, maintainability, and efficiency. Please go through the comments and provide a point-by-point response letter.
Improve the literature review. Add several pieces of research in 2019-2023.
Many of the results and conclusions of this paper are quite basic. I recommend expanding: Introduction, Conclusions and the Results sections. The aim should be to: 1) give a broader view of the literature on the topic and the current state-of-the-art; 2) clarify and discuss the novelty and the significance of the results obtained here, and compare them with those available in the literature, also including discussions on potential applications; 3) complete the manuscript with some additional, less basic results.
-
-
This manuscript introduces ProsperNN, a Python package designed for time series forecasting with four novel neural network architectures: Error Correction Neural Networks (ECNN), Historical Consistent Neural Networks (HCNN), Causal-Retro-Causal Neural Networks (CRCNN), and Fuzzy Neural Networks based on an Adaptive Neuro Fuzzy Inference System (ANFIS). The manuscript is well-structured and provides a comprehensive overview of the package's capabilities. However, several areas need improvement to enhance the paper's clarity, technical depth, and accessibility.
• Ensure the title is clear and accurately reflects the content of the paper. Consider revising "Introducing prosper nn" to "Introducing ProsperNN".
• Improve the citation format and ensure all references are complete.
• Include more figures to illustrate key points, such as model framework, model architectures and performance comparisons.
• Expand on the methodology section to provide more details on how the models are implemented and validated.
• Include a more detailed comparison with existing open-source implementations and highlight the unique contributions of ProsperNN.
• Present the results in a more structured manner, with separate sections for each model's performance.
• Add a section on future work to suggest potential extensions and improvements to the package.
• Ensure the title is clear and accurately reflects the content of the paper. Consider revising "Introducing prosper nn" to "Introducing ProsperNN".
• Improve the citation format and ensure all references are complete.
• Include more figures to illustrate key points, such as model framework, model architectures and performance comparisons.
• Expand on the methodology section to provide more details on how the models are implemented and validated.
• Include a more detailed comparison with existing open-source implementations and highlight the unique contributions of ProsperNN.
• Present the results in a more structured manner, with separate sections for each model's performance.
• Add a section on future work to suggest potential extensions and improvements to the package.
-
1. Clarity and Use of Professional English:
The manuscript is well-written with clear and professional English throughout. However, the section on neural network architectures could benefit from clearer definitions of specific technical terms upon their first appearance. This would aid in making the manuscript more accessible to non-specialists.
2. Literature References and Field Background:
The paper provides an extensive review of relevant literature and adequately situates the study within the broader field of knowledge. However, for some newly introduced concepts, such as “Fuzzy Neural Networks,” it is recommended that the author expands on the theoretical underpinnings and previous research to enhance the depth of the discussion.
3. Professional Article Structure, Figures, Tables, and Data Sharing:
The structure of the article is coherent and adheres to academic standards. Figures and tables are of high quality, relevant, and well-integrated into the text, effectively supporting the study's findings. However, the manuscript does not provide links or details on accessing the raw data, which is essential for replication and further research. The author should include detailed information on data access or provide a justification in line with the journal's data sharing policy.
4. Self-contained Presentation with Relevant Results to Hypotheses:
The manuscript is self-contained, presenting a coherent flow from hypotheses through to methodologies, results, and analysis. It avoids unnecessary subdivision of the work, which could inflate publication count. In the discussion section, the author could strengthen the impact of the findings on existing theories and practices to enhance the manuscript's applicational value.
5. Formal Results Including Definitions of Terms and Detailed Proofs:
While the manuscript defines terms and presents theorems introduced with the new neural network models, some of the more complex theorems and proofs are not detailed sufficiently. For instance, the section on Error Correction Neural Networks would benefit from a more detailed mathematical explanation and proof of the error correction mechanism. The author is advised to include comprehensive mathematical proofs in an appendix or provide additional supporting material links.
Original Primary Research and Research Question:
The manuscript presents original primary research that is within the aims and scope of the journal. The research question is well-defined and directly addresses a significant gap in the field of neural network architectures for time series forecasting. However, while the paper mentions the novelty of the introduced neural network models, it could better articulate how these specific models fill existing gaps in predictive capabilities or efficiency compared to current standards. I recommend enhancing the introduction by clearly outlining comparative advantages and situating the work within current research debates.
Rigor and Ethical Standards:
The investigation appears to be conducted rigorously and adheres to a high technical standard. The application of neural network models in time series forecasting is well-documented, and the technical aspects of the models are thoroughly described. However, the manuscript could improve by including a section on ethical considerations, especially concerning data privacy, if applicable. Given the use of potentially sensitive economic data, assurances regarding data handling, anonymization, and consent (where applicable) should be addressed.
Methodological Detail and Replicability:
The methods are detailed with a clear description of the neural network architectures implemented, including parameters and configuration details which are crucial for replication. The use of PyTorch and the provision of code via GitHub are commendable, ensuring that the methods can be replicated by other researchers. To further enhance replicability, the manuscript should provide more explicit details about the data preprocessing steps, the exact dataset used (if not proprietary), and any specific settings in the software environment that are critical for reproducing the results.
Impact, Novelty, and Replication:
The manuscript does well to establish the utility of the proposed neural network architectures in time series forecasting. However, the assessment of impact and novelty is somewhat underexplored. While the study presents new methods, it lacks a thorough comparison with existing models to clearly highlight the advancements made. Additionally, while the manuscript encourages meaningful replication, it does not provide a compelling rationale for why these specific replication studies are necessary, nor does it detail how they add value to the existing literature beyond the initial presentation of methods.
To improve, the authors should:
Provide a more detailed comparison of the proposed models against existing benchmarks in the field to demonstrate clear advances in performance, efficiency, or accuracy.
Explicitly justify the need for replication studies by discussing gaps in current methodologies that their models address, and how further replication could validate and extend these findings.
Data Robustness and Availability:
The manuscript claims that all underlying data has been provided and asserts the robustness and statistical soundness of the data used. However, there is no clear information on where this data can be accessed, nor is there confirmation that the data is stored in a discipline-specific repository that would allow for independent verification and replication by other researchers.
Suggested improvements include:
Providing explicit links to the repositories where the data is stored, or detailed information on how to access the data.
Ensuring that the data shared complies with ethical standards for data sharing, including necessary anonymization and privacy safeguards where applicable.
Conclusions and Connection to Research Question:
The conclusions of the study are well articulated and generally align with the original research questions posed. However, some claims, particularly those suggesting improvements over existing methods, lack sufficient empirical support from the data presented. This is particularly important when claims of causative relationships are made; these should be backed by robust experimental data demonstrating causality, not just correlation.
Recommendations for revision:
Tighten the conclusions to more closely reflect only what is directly supported by the results.
If causal claims are made, ensure that a proper experimental design or statistical analysis that can support such claims is clearly presented and justified.
The paper presents the prosper_nn package, which introduces four neural network architectures for time series forecasting and includes tools for sensitivity analysis and heatmap visualization. While the work is innovative and addresses a gap in available open-source models, the paper needs revisions to enhance its clarity, depth, and impact.
The explanations of the neural network architectures, particularly ECNN, HCNN, CRCNN, and fuzzy neural networks, need to be more detailed and accessible to a broader audience. Including more intuitive explanations and visual aids could help readers better understand the complex concepts. The case study section should be expanded to include more detailed analysis and comparisons with other models. Including additional datasets and scenarios could demonstrate the robustness and versatility of the proposed architectures. The paper would benefit from a more extensive benchmarking exercise, comparing the proposed architectures with a broader range of existing models across different datasets. This would help establish the relative strengths and weaknesses of the proposed methods. The authors should review the code for the prosper_nn package to ensure it follows best practices for readability, maintainability, and efficiency. Unit tests and continuous integration would also be beneficial to ensure the code's reliability. The paper should include a more thorough discussion of the implications and potential applications of the proposed architectures. The conclusion should reflect on the contributions of the work and outline directions for future applications.
no comment
no comment
n/a
All comments have been added in detail to the last section.
All comments have been added in detail to the last section.
All comments have been added in detail to the last section.
Review Report for PeerJ Computer Science
(Introducing prosper_nn - A Python package for neural network architectures)
1. Within the scope of the study, a python package containing four neural network architectures for time series forecasts, called prosper_nn, has been developed, which is shared as open source on the GitHub platform.
2. In the introduction, the development of neural networks with open source libraries and the fact that the application of the Error Correction Neural Networks architecture can be carried out with prosper_nn are basically mentioned. In this section, the literature review in the first paragraph should definitely be detailed by supporting it with a table. In addition, it is recommended to mention the importance of prosper_nn, which was developed within the scope of the study, in its use in time series forecasting problems and its differences in the literature, and its differences in the literature at the end of this introduction section.
3. In the study, the architectures, theories and implementations for Causal-Retro-Causal Neural Network, Error Correction Neural Network, Historical Consistent Neural Network and Fuzzy Neural Network, to which the proposed prosper_nn package can be applied, are sufficiently explained.
4. When the performance of Error Correction Neural Networks is compared with other Recurrent Neural Networks and the heatmaps obtained are examined in detail, it is observed that the results are appropriate and the importance of using prosper_nn can be emphasized.
5. Although the discussion and conclusion sections are also considered to be basically appropriate, it is recommended that this section be further detailed for future works.
As a result, although this study has largely proven itself in terms of its applicability to the solution of time series forecasting problems, step-by-step explanations should be made for the sections listed above in order to make its contribution to the literature clearer.
Important Note: In order to avoid any ethical problems in the future as the journal, referees and authors; In case of a possible revision process within the scope of this study, please do not add irrelevant references that are off-topic to the paper.
All text and materials provided via this peer-review history page are made available under a Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.