Relationship between Highway Geometric Characteristics and Accident Risk: A Multilayer Perceptron Model (MLP) Approach
, Bijiang Tian 1,2,*, Yuanwen Cao 3, Wenchen Yang 1,2 and Feng Zhu 4
Round 1
Reviewer 1 Report
1. As there are many machine learning models, the reasons for selecting the MLP should be specified.
2. It would be better to add a separate data section to describe the accident data as well as the basic freeway data.
3. Usually when developing the crash prediction model, the AADT should be considered as an exposure, which is not the case in this study. Please explain the reason.
4. Page 12, line 347-351: the sentence needs to be rephrased. The Nanfu highway seems to be only used for verifying the geometric alignment extraction method.
5. Please add the literatures that using MLP for road safety analysis.
6. Something seems to be missing for Figure 5(b)
Author Response
The authors sincerely appreciate the reviewers’ time and constructive comments to help us improve this manuscript. We have addressed the reviewer's comments. Please see the attached file.
Author Response File: 
Author Response.pdf
Reviewer 2 Report
This work aims to collect road design data in large-scale research and analyzes the accident risk of highway geometric alignment based on the SHAP and MLP theories.
The topic of the paper is surely of interest for this journal, even if there is little unclear in the paper. The authors should address the questions below:
1. The manuscript should be carefully proofread. Some examples of problems are as follows: “In addition, it is estimated that low- and middle-income countries account for approximately 90 % of the global total of road traffic fatalities” in line 31-32; “where X,Y,Z are global coordinate” in line 128; “in the 1950s[26]” in line 172.
2. The authors state in line 149- 150 that “when the slope exceeds ±8 %, it is considered that this is either a bridge or tunnel section, and the value was replaced by ±1 %”. Please provide the supported information or references for this conclusion.
3. The authors should provide more information of accident records. In addition, the current manuscript addresses more on data mining method. The content related to accident is relative limited. This problem should be solved as far as possible.
Author Response
The authors sincerely appreciate the reviewers’ time and constructive comments to help us improve this manuscript. We have addressed the reviewer's comments. Please see the attached file.
Author Response File: Author Response.pdf
Reviewer 3 Report
1. In the introduction section, the authors provide a brief description of road accidents and the risks associated with accidents.
2. Details about the highway's geometric characteristics and basics are missing in the introduction.
3. Details pertaining to MLP are not provided in the Introduction section. I would suggest adding discussions related to the advantages of MLP and why MLP is preferred in comparison to other state-of-the-art techniques.
4. Add your contribution with the third bulleted statement.
5. In the introduction section, the authors describe the formulation of traditional statistical models without giving citations. Provide appropriate citations.
6. A state-of-the-art literature study and comparison are missing.
7. About 23 equations were mentioned in the methodology sections, clearly describing their use in the selected scenario. The author can remove the basic equations and can focus on the main required content. The purpose of the equation with the actual link with implementation and experiment study should be added.
8. Need for equations 10,12 and 13 is not clear. Why author mention these equations?
9. Equation 11 is missing.
10. Dataset details are clearly mentioned by the author, but experimental setup details are missing.
11. Add details about the experimental study.
12. A statistical analysis is missing from the current experimental results to further justify that the proposed approach is comparable with the traditional approach in terms of those performance metrics
Author Response
The authors sincerely appreciate the reviewers’ time and constructive comments to help us improve this manuscript. We have addressed the reviewer's comments. Please see the attached file.
Author Response File: Author Response.pdf
Reviewer 4 Report
The main aim of the study is to collect road design data in large-scale research and analyzes the crash risk of highway geometric alignment. Authors used method based on satellite maps and clustering algorithms to calculate the geometric alignment of the highway plane and its longitudinal section. The crash risk of the highway geometry was analyzed based on the SHAP and MLP theories. Overall, the manuscript proposes interesting approach. However, there are several thing which should be improved before publishing.
More detailed comments are presented below:
11. First sentence in the Abstract “Highway traffic safety is the key to green and sustainable traffic development.” – In my opinion this is overstatement. Safety is of course very important, but I would not say it is the key for green traffic.
22. Row 37: main groups of factors that impact road safety are long known so I would not say that “Recent studies have shown…”
33. Introduction part gives an overview of some of the most used method for modeling road crashes. However, I’m missing better elaboration of the research gap and how this study will fill this gap.
44. Is there a reason why crash data between 2016 and now is not included in the study?
55. Last column in the Table 1 is not necessary since the collected time frame is described in the text.
66. How accurate is the calculation of flat curve length, flat curve deflection angle, longitudinal slope 357 length, and longitudinal slope gradient? Did authors tested the calculation method?
77. Discussion section is missing. Authors should connect their results with the findings from the literature and elaborate on them. Also, authors should state limitations of the study as well as direction of future studies.
Author Response
The authors sincerely appreciate the reviewers’ time and constructive comments to help us improve this manuscript. We have addressed the reviewer's comments. Please see the attached file.
Author Response File: Author Response.pdf
Round 2
Reviewer 3 Report
1. Details about the highway's geometric characteristics and basics like are still missing in the introduction.
2. Need for equations 10,12 and 13 is still not clear. Why author mention these equations?
3. A thorough proofreading is required
4. Equation 11 is till missing, there are two equations with s.no 14
5. The authors are encouraged to provide a comparative analysis of their results with that of previous research in a tabular manner
Author Response
Please see the attachment.
Author Response File: 
Author Response.docx
Reviewer 4 Report
Authors have acknowledge all my comments.
Author Response
The authors sincerely appreciate the reviewers’ time and constructive comments to help us improve this manuscript.
