Multi-Step Parking Demand Prediction Model Based on Multi-Graph Convolutional Transformer
Round 1
Reviewer 1 Report
Comments and Suggestions for AuthorsTitle: Multi-step Parking Demand Prediction Model Based on Multi Graph Convolutional Transformer
This paper (systems-3273636) presents a multi-step parking demand prediction model based on Multi-Graph Convolutional Transformer. As the number of motor vehicles in urban areas increases, the inefficient use of parking spaces has exacerbated the problem of parking shortages. Accurately predicting future parking demand can effectively improve parking space utilization and alleviate parking difficulties. This study proposes a deep learning model that captures geographical spatial features through a Multi-Graph Convolutional Network (MGCN) module and leverages a Transformer module to extract temporal feature patterns, enabling precise predictions of future multi-step parking demand. The below comments for authors’ improvements further.
1. In Section 3.2, "Parking Lot Distance Features," the authors present several calculation formulas, but the explanations for the parameters are provided only after the formulas. The authors should provide a unified explanation of all the parameters at the beginning of this section to help readers better understand the formulas.
2. In Section 4.2, regarding the GCN part, the authors should consider providing relevant pseudocode to help readers better understand the model construction process.
3. In Section 5, "Results and Analysis," some charts could be added to visualize the model's prediction results, such as comparing the prediction error distribution of different models on the same test set. When comparing models, in addition to using metrics like R², MAE, and RMSE, it might be beneficial to introduce other relevant metrics, such as Mean Absolute Percentage Error (MAPE), to provide a more comprehensive evaluation of the model's performance.
4. Still in Section 5, expanding the range and scale of the dataset, such as including parking data from different cities and various time periods, would help validate the model's generalization capability and robustness. Conducting a longer-term tracking study to assess the model's predictive performance across different seasons and events (such as large-scale activities and holidays) would also be beneficial.
5. Finally, in Figure 4, "MGCN-Transformer Model," some module names are quite vague, which affects the readers' understanding. The authors should optimize the figure further for clarity.
Comments on the Quality of English LanguageThe language level should be improved much.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Reviewer 2 Report
Comments and Suggestions for AuthorsRemarks:
1. When abbreviations are used in the paper, they should be explained at the first use. Not everywhere there are such explanations. Some abbreviations are explained only in the final part of the paper. This makes following the article a burden.
2. The paper contains different formats and sizes of writing. (Review all figures, tables, but also the areas of explication of terms in mathematical relations in the article).
3. In the text, the references are "Figure" or "Table", and in the explanations of the figures or the table headers, abbreviations are used. I do not understand this inconsistency.
4. In the structure of the text, a line of spacing is missing before or after the presentation of a figure or table, which makes it difficult to follow these elements.
5. Pay attention to the punctuation marks and spacing at the end of sentences and phrases. In some places, these are still missing.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Reviewer 3 Report
Comments and Suggestions for AuthorsThe topic of this paper is related to the topic of the journal. The researchers analyzed the parking demand prediction model based on a multi-graph convolutional transformer to address the problem of parking space utilization in cities and effectively optimize the utilization of parking spaces.
The study is original and relevant and therefore scientifically sound and is aimed at researchers and planners to improve the accuracy of parking demand prediction. The analysis shows the potential of the proposed model to achieve both high accuracy and parameter efficiency. This study fills the gap in this area of research through the input features included in the study. The input features include time series of parking demand, spatial distance features of parking lots, features of points of interest (POI) near parking lots, weather features, and vacation features. The manuscript should be better structured, it does not contain all the information necessary for the reader and it lacks a "Discussion" section, which should definitely be included in such a study and is an important part of any manuscript.
The tables and figures are informative, but the visibility of the data on the figures could be improved.
The conclusions are coherent and consistent with the arguments. The conclusions do not include information about who this study is aimed at and there are no suggestions for further research.
The references are adequate, but there is a need to reference some of the data and figures mentioned in the paper.
2 out of 25 references are from the authors of the manuscript.
Despite the positive facts, the manuscript still needs some minor and major changes and detailed explanations in preparation for publication. Once these suggestions are accepted, the manuscript will be suitable for publication.
Here are suggestions and comments:
Section 1
In paragraphs 29-31, the authors state, “According to statistics from the Chinese Ministry of Public Security, the total number of motor vehicles in the country reached 411.7 million in 2022, with 319 million being automobiles” Please provide reference for these data.
Figure 1 - Please provide reference for the figure; this figure is not mentioned in the text.
The authors state in paragraphs 48-51: “Numerous models have been proposed in past research for parking demand prediction tasks. In the early stages, mathematical models such as ARIMA models, Markov chain models, Kalman filter models, etc., were used to analyze parking demand time series for prediction tasks” Please provide references for these models.
Section 3
This section should include some of the text presented in sections 3.1-3.4.
Section 3.1
This section should contain detailed information about the district of Nanshan where the parking lots are located. What type of district is it (residential, commercial, industrial, leisure, shopping, cultural activities, …)? What are the points of interest (POI) in this area?
When was historical data on parking volume collected as historical data on parking demand (0-24, month, day, year)? Detailed information about the location of parking lots. Are they free of charge, are there parking garages? How many parking spaces per parking lot. What is the average distance between parking lot and POI?
Section 3.3
Forumla (2) has a typo. It is not “Peason” but "Pearson”. Please use the symbol of the sum instead of the letter “E”.
Section 3.4
There is no reference for Figure 2. Please explain how the data shown in the figure was collected, where and when it was collected and for which parking lot. The X-axis has no legend.
There is no reference for Figure 3. Please explain how the data shown in the figure was collected, where and when it was collected and for which parking lot. The X-axis has no legend.
Section 4.
Part of the text in Figure 4 is not legible.
Section 5.1
Figure 6 should be of better quality.
Section 5.3.1
Figure 7 X and Y axis has no legend. Please explain how the actual data was collected and what this figure represents. Please explain the relationship between the actual data and the 43 parking lots (as mentioned in section 5.1).
Section 5.3.3
The text in Figure 9 is not legible.
Section 6
This should be the "Discussion" section. Some parts of Section 5.3.2 could be mentioned here, perhaps text between paragraphs 455 and 475. Explain the different input features (spatial, weather, location, POI, driver behavior) and their impact on the model. Discuss the implications and problems that would arise if someone wanted to apply this model to another region of the world.
Once these suggestions are accepted, the manuscript will be suitable for publication.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Round 2
Reviewer 1 Report
Comments and Suggestions for AuthorsThanks for the authors' revision (systems-3273636-v2). The paper was improved a lot.
I ensured that the authors addressed the concerns raised appropriately and improved the paper. I am satisfied with the current version and would like to recommend it to be accepted for publication.
It would be much better if the authors could promote the quality of the figures like the resolution ratio.
Comments on the Quality of English LanguageThe language level is clear.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Reviewer 3 Report
Comments and Suggestions for AuthorsSection 6 should be named Discussion and conclusions.
Accept paper in present form.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf