Enhanced Linear and Vision Transformer-Based Architectures for Time Series Forecasting
Round 1
Reviewer 1 Report
Comments and Suggestions for AuthorsThe article discusses the problem of forecasting time series based on Linear and Vision Transformer-based Architectures.
The introduction outlines general approaches to time series modeling and forecasting. The development of machine learning methods, new models and approaches has appeared.
The article provides a review of the literature on deep learning methods for time series modeling. The article describes the novelty by modifying the normalization and adaptation algorithms of transformers to improve time series forecasting.
To improve the quality of presentation of research results, I recommend paying attention to the following notes:
1. Tables 1,2,3,4,5 have different styles. Authors need to make them look correct.
2. In conclusion, it is necessary to provide quantitative assessments of the quality that was achieved by the proposed approach. Section 4 has a table presenting the results, but the overall benefit is difficult to understand.
3. Section 4 shows testing models on data sets. The dataset is need to be characterized in more details (number of time series, their properties, length, etc.), if possible, links to it.
Author Response
modifications made according to Reviewer 1 comments:
Reviewer 1 Comment 1:
- Tables 1,2,3,4,5 have different styles. Authors need to make them look correct.
Modification made: We have corrected the style to make it uniform font and appearance for all tables in the paper.
Reviewer 1 Comment 2:
- In conclusion, it is necessary to provide quantitative assessments of the quality that was achieved by the proposed approach. Section 4 has a table presenting the results, but the overall benefit is difficult to understand.
Modification made: We have added a new table (Table 4) which provides quantitative improvements over two existing recent approaches. Further explanation and the quantitative benefits are explained by a paragraph in section 4, as well as conclusions has been enhanced to summarize the quantitative improvements..
Reviewer 1 Comment 3:
- Section 4 shows testing models on data sets. The dataset is need to be characterized in more details (number of time series, their properties, length, etc.), if possible, links to it.
We have also added a row in Table 2 which provides more detail on the datasets used and the link to datasets is provided at the end of the paper.
Author Response File: 
Author Response.pdf
Reviewer 2 Report
Comments and Suggestions for AuthorsThis paper delves into linear and vision Transformer-based architectures for time series forecasting, introducing methodological enhancements to existing methods and in turn applying them to existing datasets. The authors demonstrate a thorough knowledge of the topic, and the paper addresses an interesting and timely research theme.
However, the paper presents important issues, particularly in its structure and the lack of a clear description of its scope and contributions to the literature. Consequently, the paper requires major revisions to be considered for publication in MDPI’s ‘’Big Data and Cognitive Computing’’ journal. Since the journal only allows major revisions in cases where control is missing from some experiments, I recommend minor revisions through the review system. However, before accepting the paper, the editor must ensure my comments (detailed below) have been addressed.
1) The abstract must be revised to be more concise and structured. It should briefly mention the research gap it aims to address, and outline the scope of the paper, highlighting the key results and contributions. The literature review must be excluded from the abstract, while the GitHub link should be relocated to a more appropriate place.
2) Proper references are missing from the paper in several instances (e.g., line 39), particularly where specific claims are made (e.g., line 51). Proper referencing should be confirmed throughout the paper.
3) In the Introduction, the authors don’t outline the scope of the paper, the research gap it aims to address, and its contributions to the existing body of literature. These elements should be clearly described in Section 1, rather than first appearing in Section 3, where they are still not clearly explained.
4) In section 3, the rationale behind each methodological component should be clearly explained. Currently, each step appears unexpectedly without clear reasoning or context.
5) In the Results section, large tables could be moved to an appendix, with the main body discussing their implications. The results should not merely present the data from tables and figures, but crucially should also discuss the underlying reasons for these outcomes and how they relate to the methodological enhancements made. The same logic should be adopted in the Discussion section. As the authors claim that the ‘’model surpasses most established baseline methods in majority of the test cases’’, they should disentangle the drivers of the results and present a comprehensive and deep discussion of them in relation to the existing body of literature in the field.
6) Similarly, the Conclusions section must provide deeper explanations behind each statement made. This section should also address the limitations of the study and describe future research directions more broadly, rather than framing them as the authors’ personal agenda as currently happens.
7) ‘’Although’’ and ‘’however’’ are frequently misused (e.g., line 14, line 20), which requires revision.
8) The list of references should be expanded to include a wider body of relevant literature.
Comments on the Quality of English Language
The paper is generally well-written. However, it requires revisions to improve its readability and rectify existing errors.
Author Response
Summary of modifications made according to Reviewer 2 comments:
Reviewer 2 Comment 1:
- The abstract must be revised to be more concise and structured. It should briefly mention the research gap it aims to address, and outline the scope of the paper, highlighting the key results and contributions. The literature review must be excluded from the abstract, while the GitHub link should be relocated to a more appropriate place.
Modifications made: We have revised the abstract and made it more concise. The Github link to the code has been moved at the end of the paper.
Reviewer 2 Comment 2:
- Proper references are missing from the paper in several instances (e.g., line 39), particularly where specific claims are made (e.g., line 51). Proper referencing should be confirmed throughout the paper.
Modifications made: We have carefully gone over the paper and added references in a few places where we had missed these.
Reviewer 2 Comment 3:
- In the Introduction, the authors don’t outline the scope of the paper, the research gap it aims to address, and its contributions to the existing body of literature. These elements should be clearly described in Section 1, rather than first appearing inSection 3, where they are still not clearly explained.
Modifications made: We have added paragrphs at the end of section 1 and in section 3 to address the research problem that we are addressing and improving upon. The added paragraph in section 3 provides a better continuity in the different subsections.
Reviewer 2 Comment 4:
- In section 3, the rationale behind each methodological component should be clearly explained. Currently, each step appears unexpectedly without clear reasoning or context.
Modifications made: We have added a paragraph in section 3 to provide a better motivation and continuity to the next subsection.
Reviewer 2 Comment 5:
- In the Results section, large tables could be moved to an appendix, with the main body discussing their implications. The results should not merely present the data from tables and figures, but crucially should also discuss the underlying reasons for these outcomes and how they relate to the methodological enhancements made. The same logic should be adopted in the Discussion section. As the authors claim that the ‘’model surpasses most established baseline methods in majority of the test cases’’, they should disentangle the drivers of the results and present a comprehensive and deep discussion of them in relation to the existing body of literature in the field.
Modifications Made: We have added an extra table for quantitative comparisons with two most recent models. Related explanations of the results are further provided by an additional paragraph in the results section. We believe that the tables pertaining to the results are important part of the main paper so in our opinion, we prefer to keep these in the main paper rather than moving them to the appendix. If the editor or the reviewer strongly feel that we should move any of the tables to the appendix, we will oblige.
Reviewer 2 Comment 6:
- Similarly, the Conclusions section must provide deeper explanations behind each statement made. This section should also address the limitations of the study and describe future research directions more broadly, rather than framing them as the authors’ personal agenda as currently happens.
Modifications made: Conclusions and future research sections have been improved.
Reviewer 2 Comment 7:
- ‘’Although’’ and ‘’however’’ are frequently misused (e.g., line 14, line 20), which requires revision.
Modifications made: These have been modified as we revised the abstract.
Reviewer 2 Comment 8:
- The list of references should be expanded to include a wider body of relevant literature.
Modifications Made: We have corrected a duplication in one of the references and added a few more recent references.
Author Response File: Author Response.pdf
