Lashing Force Prediction Model with Multimodal Deep Learning and AutoML for Stowage Planning Automation in Containerships

Round 1

Reviewer 1 Report

1. In Multimodal approach, using ANN, CNN is fine but poor justification provided for using RNN as a separate model. In fact with the justification you have provided in line [175-178] it should be used as a feature in all models. 

2. Deep neural networks architecture for all models are not shown graphically, model summary output from Keras does not qualify. It should be drawn and explained. 

3. Best model parameters does not show epochs if early stopping is used. It is recommended to use early stopping when using such a high number for epochs such as 1000. 

3. Grammar check is highly required in many sections, give a consideration over consistency. 

4. Conclusion is not explained well. it should not contain methodology and technology usage, it should contain future aspects and practical feasibility.

Please check markup comments in attached pdf. 

Thank you

Comments for author File: Comments.pdf

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

1- The literature review part needs major revision. There are a couple of papers related to the stowage planning problem from recent years, but neither of them has been referred. Near half of the references are from before the last decade. The importance of the problem has not been explained very well. 

for example, these are just two of them:

"Matheuristics for slot planning of container vessel bays(2020)"

"A Rule-based Greedy Algorithm to Solve Stowage Planning Problem(2018)"

"Route-Specific Container Stowage(2013)"

- Adding more references to previous studies of using ML in stowage planning and lashing are needed. If this is the first study, it would be better to mention or if it is not, add more references to be more clear the contribution. 

The author claims that " there is little to none study on lashing forces, taken into consideration for stowage planning automation..." . There are some studies about the lashing forces and he(she) needs to reference them and explain the process they used for improving the calculation of the lashing force, even if they are not automating it.

2- The quality of the pictures must be improved (Especially Fig 5 and Fig 7). More explanation is needed for the figures especially fig 4 and explain what is going on there.

3- The contribution is not clear enough. In the contribution, the author needs to exactly mention which problem they are solving and what is the benefit of their works.

4- In the " Lashing Force Prediction with Multimodal Deep Learning" part, Describe the proposed process more clear and cohesive.

5- In the part " In severe sea conditions as well as in case of improper container stowing and container overweight, these forces may become excessive ....." it would be better to exemplify some real-world issues to make the problem more clear to the readers and show the importance of the proposed idea.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 3 Report

The subject of this paper is of great interest to readers but the presentation needs much work.  Graph axes are not labelled, sentences roll on over more than 4 lines where multiple sentences should be used, singular nouns are used with plural verbs almost constantly.  It is recommended that the paper be re-edited by an editor who has English as a first language.  The selection of the values in the Best Model result is not clear.  Figs 4, 6 & 7 need to be explained more fully as to significance.  Section 2.2.5 introduces ANN, RNN and CNN but are not mentioned in the Abstract nor the Introduction nor in Section 1.4 where it would be expected that these would be mentioned as something coming up later in the paper.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 4 Report

Dear authors

I have the following suggestions and conclusions regarding to your manuscript.

1- Please define TEU?

2- The data used in this research is not clearly explained. Please briefly explain the data used in this research which should include data source, quality, accuracy etc.

3-The overall research is not designed well, the whole paper looks like a report.

4- Where are the discussion of the results?

5- where are some comparisons with existing methods?

6- Conclusions section need to be improved.

In the manuscript, only model results are reported with a sound scientific explanation.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 2 Report

The literature review still needs some works. The title is about lashing force with DQN but in the literature review, no works are referencing calculating lashing force. There are many papers from the last decade about stowage planning and different methods used, those can be included in the literature review.

The issue must be addressed in the literature review part, but in this work, this part looks artificial.

Author Response

Thanks for valuable feedback.

I have updated literature review according to your feedback.

Please have a look.

Reviewer 3 Report

Significant improvement noted.

Author Response

Thank you so much for your valuable review in the first round and compliment this round for improvement.

I will check the English grammar as you marked in manuscript

Reviewer 4 Report

The authors have made significant changes to the current version of manuscripts by addressing the comments. I would recommend publishing in the present form.

Author Response

Thank you for your kind review in the first round and recommendation for publishing.