Automatic Identification System-Based Prediction of Tanker and Cargo Estimated Time of Arrival in Narrow Waterways
Round 1
Reviewer 1 Report
Comments and Suggestions for Authors1. In maritime transportation, the use of machine learning algorithms for predicting Estimated Time of Arrival (ETA) has become quite common, and this alone does not serve as the novelty of this article.
2. The article provides a detailed introduction to relevant methods such as XGBoost, but falls short in clearly explaining how to apply these algorithms to achieve the prediction of Estimated Time of Arrival (ETA).
3. The influencing factors of ship navigation come from various aspects. For the prediction of Estimated Time of Arrival (ETA), it is necessary to comprehensively consider both temporal and spatial factors, taking into account the influences of multiple factors simultaneously.
4. In section 5.3 of the article, it is mentioned that the model demonstrates greater accuracy for shorter travel durations. However, for longer trips (exceeding 180 minutes), the figures illustrate that travel time estimates become more dispersed around the average, indicating a decrease in accuracy for these extended journeys. Given that sea transportation journeys for ships are often of substantial duration, if the majority of ship travel times exceed 3 hours, how can the superiority of this method be demonstrated?
Author Response
Please see the attachment
Author Response File: 
Author Response.pdf
Reviewer 2 Report
Comments and Suggestions for AuthorsDear authors,
Congrats to an interesting piece of work. I have read your manuscript and found the work interesting and valuable. However, there are a few shortcomings that need amendment.
Your manuscript very much focusses on the development and implementation of algorithms and their application for the prediction of ETAs in a specific use case.
The reviewer finds the problem description too general. Yes, ETA is important for planning but at no place of your text you explain, which of the ETA is meant. It is not clear what “exact ETA”, “accurate ETA” or “precise estimation” is meant and to what extend the improvement will solve what real problem of resource planning.
On the other hand, you only mentioned very few factors influencing the transit times of a ship to the berth at a terminal. Did you really analyze if the parameter you mentioned are the most important and relevant ones?
Although your title mention ETA in narrow fairway, there is no explanation what you consider a narrow fairway. Is it a definition from local administrations? Is it related to COLREGs? Please clarify!
Explanations of AIS (history and technical content are partly incomplete or even just faulty and not sufficient. For instance, you are talking “of extracting useful information from historical and real time data” (p5 l191). The reviewer recommends van der Heijden, 2020 and documents of IALA to amend this paragraph!
Are there useless AIS information? How do you differ historical AIS data from realtime data? Does it mean you are not only using AIS information? Please clarify!
Moreover, source [51] is incomplete or not a reference!
Figure 1 is not very illustrative! The reviewer rather misses a picture of the area for which you did the experiment! Figure 3 missing information about the scale! If it is a principle sketch, an explanation about the dimensions of the segment, corridor is needed! Finally, title of figure 4 is entitled that is representing a complete trip. However, in terms of a vessels passage and the original meaning of an ETA, readers with maritime background would expect to see the destination at the berth as a complete trip from entry into the area to the final destination. So far it seems you are considering only ETA just as the passing times at artificially estimated waypoints of a vessel. Please clarify!
Although you are focusing on prediction of ETA, it could be supportive to also show how your extracted route network complies with the real routes.
In your conclusion, the sentence ending “… and finally prediction vessel arrival time between to location in maritime route.” (p14 line 469) is unclear.
You are stating that your tool “While it showed higher precision for shorter trips, its predictions for longer trips (over 180 minutes) displayed a wider dispersion around the average values”. If you did not expect this, what is the relevance of this statement? Does it mean your model is insufficient for a more comprehensive (complete) ETA prediction from the entry into a fairway up to berthing (only there the port operation really begins)? If so why and for what purpose on will need the “shorter-term” ETA prediction?
The reviewer also misses any reference to ongoing developments in exchanging route plans via AIS and existing arrival planning systems of ports, VTSs or VCCs and the relation of your research to such systems. This would add further value to your good draft manuscript and upgrade it to a very good or even excellent manuscript.
I am looking forward to read about your further work and wish you all the best for this.
Yours sincerely
Comments on the Quality of English LanguageEnglish expression is very well, there are only minor errors (typos) that quickly will be revised (e.g. p.4 l179ff passive tense, p10, 375 missing link, p 11: Figure 5 title of diagrams).
Author Response
Please see the attachment
Author Response File: Author Response.pdf
Reviewer 3 Report
Comments and Suggestions for AuthorsWell done!
1. The abstract provides a concise overview of the study, emphasizing the importance of accurately predicting the Estimated Time of Arrival (ETA) for vessels in maritime logistics.
2.The use of machine learning, specifically an XGBoost model, for ETA prediction using historical Automatic Identification System (AIS) data from 2018 to 2020 is highlighted.
3. The research framework, including data preprocessing and feature engineering, is well-defined.
4. The validation using AIS data from the Port of Houston demonstrates promising results, with a Mean Absolute Percentage Error (MAPE) of 5%.
5. The abstract concludes by suggesting potential benefits in terms of reduced complexity and increased generalizability in maritime ETA predictions.
6. The introduction effectively contextualizes the significance of shipping in global trade, emphasizing the critical role of efficient port operations and the impact of shipping delays on port efficiency.
7. The paper underscores the importance of accurate ETA predictions and highlights the existing gap in research for shipping systems compared to land-based travel time estimation.
8. The integration of machine learning for ETA prediction in maritime routes is well-motivated and distinct from traditional methods used in city planning or land-based transportation.
9. The section establishes the central role of the proposed machine learning framework in addressing the challenges of estimating travel time for vessels.
10. A comprehensive review of related works is presented, including discussions on ETA prediction in air traffic control, road transportation, and maritime transportation.
11. The paper acknowledges the lack of studies on vessel arrival times in ports using historical tracking data and identifies the unreliability of ETA from AIS messages.
12. The incorporation of machine learning algorithms for qualitative estimation of vessel ETA is appropriately emphasized, showcasing the relevance of such approaches in maritime logistics.
13. The section provides an insightful exploration of various path-finding algorithms and their applications in estimating vessel arrival times.
14. Notable algorithms like Dijkstra, A*, and data-driven methods are discussed, with examples of studies using these algorithms to address specific challenges in maritime traffic control.
15. The paper successfully introduces the concept of artificial intelligence and machine learning in the context of maritime logistics.
16. Different machine learning applications in road and waterway scenarios are discussed, demonstrating the versatility of machine learning in addressing travel time prediction challenges.
17. The section provides a detailed overview of machine learning applications for road transportation, emphasizing the use of GPS traces and live AVL data for travel time prediction.
18. Examples of studies employing neural networks, recurrent neural networks, and novel combinations of optimization algorithms are discussed, showcasing the diversity of approaches in road applications.
19. The application of machine learning, particularly artificial neural networks (ANNs), for solving shipping-related issues such as container flow forecasting and navigational behavior prediction is well-explained.
20. Notable studies using data mining methods, path-finding algorithms, and various machine learning models to predict vessel arrival times at specific ports are comprehensively reviewed.
21. The paper effectively concludes by summarizing the key contributions, emphasizing the efficacy of the proposed machine learning framework for ETA prediction in maritime logistics.
22. The structured approach in subsequent sections, promising results from experimentation, and the identification of potential future research directions provide a solid foundation for concluding the paper.
In summary, the research paper provides a thorough exploration of the proposed machine learning framework for maritime ETA prediction, adequately addressing the gap in literature, and showcasing promising results through comprehensive validation. The incorporation of related works and the contextualization of the study within the broader field of transportation and logistics enhance the paper's scientific merit.
To conclude, the paper is a rare example of well-prepared paper ready for publications.
Author Response
Thank you to the reviewer for the positive feedback. Your observations and suggestions have greatly helped improve our work. We appreciate the time and effort you put into reviewing our manuscript. Your supportive comments motivate us to continue our research with more dedication. Thank you again for your valuable contribution.
Round 2
Reviewer 1 Report
Comments and Suggestions for AuthorsIn its current form, I think the paper has no further issues.
Reviewer 2 Report
Comments and Suggestions for AuthorsDear authors,
thanks for your cooperation and the amendments done to improve the manuscript.
Your additions provide clarification to readers and highlights the fundamental academic approach on this study. For practioners there might remain the question whether there is a real need of AI algorithms to predict passage times at waypoints.
The reviewer prefer to use the term "ETA" only when talking of arrival at final destination, e.g. berth!
Of course, the reviewer wishes you all the best for your further work.
Sincerely
Comments on the Quality of English Languagen/a
