Application of Regression Analysis Using Broad Learning System for Time-Series Forecast of Ship Fuel Consumption
Round 1
Reviewer 1 Report
The article compares regression-based methods for predicting fuel consumption on ships. It is a clear and well-explained methodology but the following points should be corrected:
-- The title of the article is not correct as it implies that a new methodology is proposed. However, this article is merely a comparison of regression methods for a specific problem.
-- There are a multitude of poorly linked references in the paper that make it impossible to verify what is stated in the text.
-- Given that the BLS is a method based on a large intermediate layer with random interconnections, it would be convenient to compare the results with a similar technique such as ELMs.
-- The article does not specify at any point how the separation of the data in Training / Validation / Test is carried out, nor what kind of techniques are used to avoid overtraining.
-- The authors should clearly indicate what is the difference between this article and reference 29 as it seems that this article is an extension of the same, it should be clearly indicated at least in response to the reviewers how this article differs from the conference version.
Author Response
Thank you very much time and effort on our manuscript. According to the comments and suggestions, the manuscript has been completely revised. The detailed responses and revisions were listed in the attachment. We also requested English editing service to polish the language of our manuscript. Since it will take a few days to finish the English editing and polishing, we will attach the certification in final version.
Author Response File: 
Author Response.docx
Reviewer 2 Report
This paper presents a time-series forecast model of ship fuel consumption based on a regression analysis using a broad learning system (BLS). They compared the mean absolute error (MAE) and computational time performance of the proposed algorithm with autoregressive integrated moving average model with exogenous inputs (ARIMAX), support vector regression (SVR), recurrent neural network (RNN) and long short-term memory network (LSTM). In addition, this paper includes experiments to find the optimal parameters of each method and compare the obtained the MAE of each method when using the optimal parameters. The structure of this article is sufficient, and the writing is clear. There are some revisions as follows:
Authors needed to describe the scientific contribution better and clarify what is new in the proposed algorithm.
There are many acronyms and parameters: I suggest authors add a list of abbreviations and parameters at the beginning of the manuscript.
At the end of Section 1 (Introduction and literature review), suggest to add a table summarizing related literature review to show how this study differs from other work. In addition, a highlight with bullet points showing a list of paper contributions should be added.
There are a number of acronyms and parameters: I suggest adding a list of abbreviations and parameters at the beginning of the manuscript.
Citation numbers are missing from page 9 onwards. Please verify.
In Figure 7-10, In my opinion, the results of this experiment should be presented in a table showing the METF and MEA values, which would better illustrate the differences in the efficacy of each method. If the authors want to offer a graph, it should also explain the trend of each method's forecast value. Please explain the meaning of METF.
Sections 4.3 and 4.4 should be merged into one section.
In tables 6 and 7, the forecasting time performance of each method should be further described.
The conclusion section proves to be too short. This should also collect a summary of the problem, a brief discussion of the results and the future steps to cover the weak points of the current method. Please complete the final section.
Author Response
Thank you very much time and effort on our manuscript. According to the comments and suggestions, the manuscript has been completely revised. The detailed responses and revisions were listed in the attachment. We also requested English editing service to polish the language of our manuscript. Since it will take a few days to finish the English editing and polishing, we will attach the certification in final version.
Author Response File: Author Response.docx
Reviewer 3 Report
The paper is interesting and well-written.
Authors conduct research about forecasting ship fuel consumption. They build a time-series forecast model of ship fuel consumption based on a novel regression analysis using a broad learning system. The proposed model is verified with experiments, using two types of ships.
The manuscript is presented in a well-structured manner. The authors use an appropriate experimental design. The figures and images are clear and illustrate the presented results. The cited references are mostly recent publications.
I have minor observations:
* Authors should add descriptions about the used model variables.
* It is good to add some citations for the broad learning system (BLS).
* There is text "Error! Reference source not found." in lines 262, 297, 303, 306, 309, 311, 312, 362, 364.
* There are small grammatical errors. The paper should be proofread.
Author Response
Thank you very much time and effort on our manuscript. According to the comments and suggestions, the manuscript has been completely revised. The detailed responses and revisions were listed in the attachment. We also requested English editing service to polish the language of our manuscript. Since it will take a few days to finish the English editing and polishing, we will attach the certification in final version.
Author Response File: Author Response.docx
Round 2
Reviewer 1 Report
All the issues have been covered so it can be accepted
