Traffic Flow Online Prediction Based on a Generative Adversarial Network with Multi-Source Data

Round 1

Reviewer 1 Report

Dear Authors,

 

The paper is very well structured. It establishes an online Generative Adversarial Network (GAN) by combining Bi-Directional Long Short-Term Memory (BiLSTM) and Convolutional Neural Network (CNN) as the generative and discriminative model.

 

The article is consistent with the lines of the journal and the topic is of interest to readers. The paper give an interesting contribution to the debate on the subject in question.

 

However, the paper it might still require some minor improvements:

 

• The Abstract should not be longer than 200 words (at the time is 223 words)
• Abstract contain main obligatory elements (according to the instruction to the authors): Background of the research, Methods; Results; Conclusion).The research methodology, application of appropriate statistical techniques are all adequate for the analysis conducted.
• List of Keywords is appropriate.
• The paper reference list is satisfied citied in the paper.
• After every chapter name (title) should be some entrance sentence. It is not appropriate to have chapter title follow with chapter subtitle (see chapter title 3 which is followed directly with subtitle 3.1 without any description between them, or 5.5 and 5.5.1.)
• In the paper is often used “study” which should be changed with word “paper”
• The paper should be written in third face singular. Expression like “we present” should be corrected in “paper presents…”
• The introduction chapter presented problem/challenges, which will be discussed.
• First sentences in chapter 3.1 and 3.2 are the same.
• The authors should carefully read the paper and correct minor writing mistakes (missing dot at the end of the sentences, spaces between words etc.)
• Title of chapter 5.5.3 should be on other page. It is not appropriate to have only name of the chapter at the end of the page.
• The results of the research are authors explained in detail.

Author Response

Thank you for your thoughtful consideration, feedback, and insightful questions regarding our manuscript entitled “Traffic Flow Online Prediction Based on Generative Adversarial Network with Multi-Source Data”. In addressing the reviewer comments, we have extensively revised the paper, and we believe that it has been strengthened and improved as a result. We thank you for coordinating the reviews for this paper, and we hope that you will pursue the revised paper’s publication in the Sustainability (ISSN 2071-1050). A point-by-point response to the your comments has been finished. Please see the attachment.

Author Response File: Author Response.docx

Reviewer 2 Report

Based on deep learning model LSTM and CNN, the authors have proposed a Generating Adversarial Network for traffic flow framework with generator BiLSTM and discriminant CNN to predict traffic flows in lanes during fixed intervals. By comparison with four baseline models, including Autoregressive Integrated Moving Average (ARIMA), BiLSTM, Generating Adversarial Network for traffic flow (GAN-TF), and Generating Adversarial Network for non-signal traffic (GAN-NST), the proposed model generates more accurate and stable flows in the conducted experiments.

As for the general writing, the topic is interesting and the conclusion makes sense. Moreover, the manuscript is very well structured and the main purpose of this research is clear. However，the quality of this manuscript has been dramatically reduced by the following issues.

1. In the end of section ‘2. Literature review’, three key problems of the relevant researches published so far have been summarized: (1) How to find the periodicity and trend of flows in time domain? (2) How to describe the local trend raised by special cases with steady accuracy? and (3) How to maintain continuous self-learning ability in time domain? After such the summarizations, it is better to provide possible solutions to solve these problems in the end of this section.
2. In this manuscript, the intersection of Hongzehu Road and Qingnian Road in Suqian City, Jiangsu Province, China has been selected as the test area. The interval length of observed traffic flow was 15 minutes, and time domain was from October 26th, 2016, to May 9th, 2017 (totally 18,816 intervals of 196 days). From the sample data of one intersection are not enough to verify the usability of the proposed model. In other words, more sample data in various test areas should be considered.

Given both of the issues mentioned above, I recommend a general evaluation between ‘major revision’ and ‘minor revision’.

Author Response

Thank you for your thoughtful consideration, feedback, and insightful questions regarding our manuscript entitled “Traffic Flow Online Prediction Based on Generative Adversarial Network with Multi-Source Data”. In addressing the reviewer comments, we have extensively revised the paper, and we believe that it has been strengthened and improved as a result. We thank you for coordinating the reviews for this paper, and we hope that you will pursue the revised paper’s publication in the Sustainability (ISSN 2071-1050). A point-by-point response to the your comments has been finished. Please see the attachment.

Author Response File: Author Response.docx

Reviewer 3 Report

1. The main objectives of the research are defined at the introduction of the study. The authors described the study problem and research questions, the importance of the study, and the hypotheses as well.
2. Literature review covers the most important and relevant international literature sources in an appropriate structure. The literature sources are highly acceptable and most of the relevant literature sources are used, the in-text citations are used well, citation style is correct. I suggest adding this paper to reference: Izonin I., Tkachenko R., Fedushko S., Koziy D., Zub K., Vovk O. (2021) RBF-Based Input Doubling Method for Small Medical Data Processing. Lecture Notes on Data Engineering and Communications Technologies, vol 82. Springer, Cham, 2021. pp 23-31. https://doi.org/10.1007/978-3-030-80475-6_3

3. All the tables and figures are clear, understandable, and relevant, sources are indicated in each case well.
4. The authors have completed the necessary evaluations. Conclusions and recommendations are well structured, those are in relevance with the analysis and discussion. Conclusions are suitable for gaining new results and initiating further or new research. The new results are drawn up in an understandable way.
5. In materials and methods are a good and comprehensive overview of the topic, based on a wide range of literature. The methodological contains a correct description of the methods applied, is well documented and supported.
6. The conclusions show that the authors have good and deep knowledge of the topic. 

Author Response

Thank you for your thoughtful consideration, feedback, and insightful questions regarding our manuscript entitled “Traffic Flow Online Prediction Based on Generative Adversarial Network with Multi-Source Data”. In addressing the reviewer comments, we have extensively revised the paper, and we believe that it has been strengthened and improved as a result. We thank you for coordinating the reviews for this paper, and we hope that you will pursue the revised paper’s publication in the Sustainability (ISSN 2071-1050). A point-by-point response to the your comments has been finished. Please see the attachment.

Author Response File: Author Response.docx

Round 2

Reviewer 2 Report

the manuscript has been well refined.