Short-Term Traffic Forecasting: An LSTM Network for Spatial-Temporal Speed Prediction
Round 1
Reviewer 1 Report
The presented article deals with road traffic forecasting with focus on LSTM networks for speed prediction.
The methods could be described in more detail. However, I understand that there could be a significant increase in the number of pages.
The data were measured in 2016. Are there any newer data for analysis?
I recommend checking the manuscript of the article with native English speakers, or using the MDPI English editing service.
Author Response
Please see the attachment.
Author Response File: Author Response.docx
Reviewer 2 Report
Here are my comments:
First about your data. LSTM model is capable of detecting the special events. If your data shows special events such as an accident or people around an stadium, do you think your result would differ? Also, what about having minor roads instead of the major arterial ones? Could you add them all as an input?
Also, please describe the effect of changing your parameters (LSTM inputs) in the outcomes.
The second question that challenges this study is that the LSTM code is available online for prediction. An online website has the code in R for predicting something in space that can be easily translated into this application. Also, a recently published paper entitled Travel speed prediction based on learning methods for home delivery used a comprehensive analysis on this method. So, I questioned your contributions.
Author Response
Please see the attachment.
Author Response File: Author Response.docx
Reviewer 3 Report
In the paper different methodologies for short-term traffic forecast are described and compared on the same example. Possibility to implement same models at different location is analyzed.
Advantage of the experiment is that is based on real data, collected on field.
The Introduction is well organized - objectives of the paper are clear, literature review gives good inside of the present state in the field.
Data collection: which type of equipment was used, try to make one figure from figures 5. and 6. because they give very similar information
The results of calculations are presented in detail but it will be good to include Discussion and make some more general discussion on used methodologies as well as compare it to previous similar analyzes.
- line 5-9: add affiliations
- line 17 - explain RNN
Author Response
Please see the attachment.
Author Response File: Author Response.docx
Reviewer 4 Report
This draft is showing the potentials of using LSTM for traffic speed prediction.
The topic is significant and has practical meaning. However, the methodology is not very novel.
My detailed comments are as below:
- In the abstract, when an abbreviation appears, the full phrase should be presented.
- Figures 5 and 6 are repetitive. They can be generalized into one, depicting the model. Part 4 can be moved up before Data Collection.
- Data Collection + Model Evaluation can be a case study.
- Please increase the figure quality.
- MAPE is already a percentage error. Not sure what the accuracy means.
- Besides MAPE, maybe add MAE to give an intuitive sense of the speed errors.
Author Response
Please see the attachment.
Author Response File: Author Response.docx
Reviewer 5 Report
The paper suggests models for spatial and temporal predictions of traffic conditions in different locations of a particular road in Melbourne, Australia.
The English of the paper must be improved. E.g. in the first paragraph: "policy makers and decision maker" should be changed to "policy makers and decision makers". Also, "using variety of theoretical models" should be changed to "using a variety of theoretical models" and many more grammar and spelling mistakes.
Another issue in this paper is ignoring the new looming technology of autonomous vehicles. In the second section they give motivation for speed prediction, which is important, but they ignore speed of autonomous vehicles. In Y. Wiseman, "Autonomous Vehicles", Encyclopedia of Information Science and Technology, Fifth Edition, Vol. 1, Chapter 1, pp. 1-11, 2020, available at: https://u.cs.biu.ac.il/~wiseman/Autonomous-Vehicles-Encyclopedia.pdf the author writes "Likewise, Autonomous vehicles will be able to coordinate their way of driving, so they can move in platoon with fixed intervals by employing Lidar and ultrasonic rangefinders with the intention of distance reduction between them, without any risk of collisions. These platoons will reduce traffic congestion on roads, will increase the energy efficiency of driving and will reduce the air pollution generated by vehicles.". So, I would encourage the authors to cite this encyclopedia entry and explain that the speed of autonomous vehicles in platoon will be equal.
Author Response
Please see the attachment.
Author Response File: Author Response.docx
Round 2
Reviewer 2 Report
Dear Authors,
You did not convince me about your contribution as I stated that the code is available online and shared a published paper that did a similar study.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Reviewer 4 Report
Thank authors for the efforts! Please find my comments below.
- Figure 1 can be even more generalized, leave out the road names and specific distances.
- Please improve the quality of Figure 2.
- Figures 7 and 9 (missing title) should be different types of charts since there's no connection between adjacent data points.
Author Response
Please see the attachment.
Author Response File: Author Response.docx
Round 3
Reviewer 2 Report
Coding is not a contribution. A contribution defined as the goal of the study.