Multi-Target Regression Based on Multi-Layer Sparse Structure and Its Application in Warships Scheduled Maintenance Cost Prediction

Round 1

Reviewer 1 Report

This is a study to predict cost for ship equipment grade maintenance. 

1. Consumption of ship's supplies depends on the length of the voyage and the number of port visits. There are far more irregular ships than regular ports, and none of these details are available in this study.

2. Since it is a regression, an appropriate verification method is required.

 

3. No uniqueness was found in the prediction algorithm. If so, novelty should be recognized in the field of application of this study. This can be verified through the feature sets in Table 3. I can't see the rationale and logic of this distinction anywhere. Also, in my personal opinion, the question arises as to what the significance of such a prediction of a ship's equipment cost.

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 2 Report

Dear authors, your research is very interesting and may contribute to address complex problem of cost prediction of ship equipment. Your idea is new and relates sophisticated mathematical modelling with challenging problem of the cost prediction. However, your manuscript requires significant modifications in order to convey your research in a clear manner to the reader. Presently, your work cannot be reproduced independently by other researchers and professionals. Thus, the readership presently has a limited access to results of your work. Hence, I suggest following improvements:

1. You extensively use word 'subitem'. I am not sure if this is the right choice. In English dictionary I found a definition stating that it stands for an item that is less important.  I suggest that you use some alternative phrase like unit maintenance cost, or maybe better, equipment-specific cost. Or you may consider some other formulation - later on you refer to 'instance cost'.

2. Please, include 'space' between text and reference throughout the manuscript. E.g. ... as gray theory [2] instead of as grey theory[2]. A some places you include . before square brackets. Please correct.

3. Line 88 regression algorithm instead of regressor algorithm.

4. Present literature review is focused only to different algorithms. However, since you are dealing with cost predictions in the case of shipbuilding industry, you have to significantly extend the reviewed literature. Therefore, please, include literature review related to existing methods of cost predictions in shipbuilding. Also, additionally, please, include aspects of maintenance strategy (interventive, preventive, ...), maintenance planning, types of maintained equipment (cargo equipment, propulsion systems, wetted surface, energy systems, safety systems, manouvering systems, ...), discuss role of maintenance shipyards and classification societies, prescribed maintenance schedules defined by class societies.

5. All equations must be involved better in the text. This means that a reader must understand them. Therefore, please explain all symbols immediately after their first appearance in the text. E.g. in Eq. (1) you miss to introduce N and labda. And so on. Also, you must hint on the physical background of your equations by relating them with the problem you are considering. E.g. hod does a content of matrix [W] relate to problem you are addressing.

6. Lines 242-244 - a sentence need clarification. It starts with 'since'?? Also, in line 245 you have . before eq. (10). Please, clarify.

7. Eqs. (18) and (19) require font adjustments.

8. Algorithm 1 - please, denote matrices in bold style.

9. Examples: Table 1. brings ne data, but it is not clear what does it mean (datasets, samples - - of what, input - what kind of input, ...) Please clarify it so that a reader can understand its content and repeat the calculation using data in Tab 1.

10. I think a ref is missing after Bonferroni-Dunn test.

11. At the end of Section 4.2 a discussion on CPU requirements should be included regarding different approaches. How does your algorithm stand in that context?

12. Table 3 is not referred in the text. Why did you choose that specific data presented in Table 3 - please explain in the text. Why do you consider warships? Why do you omit merchant shipy? What is the source of your data on equipment maintenance?

13. What is the source of data presented in Tab 4? Please, discuss maintenance grades.

14. A reader can hardly understand Figs 3 and 4 since input data is not well presented. 

English grammar issues in lines: 9-12, 47-50, 304-306.

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 3 Report

This paper presents Multi-target regression to predict cost of ship equipment grade maintenance. I have some comments to be considered during revision as follows:

1. What is GMSE?

2. What is Mulan?

3. Could you please insert refs of the dataset?

4. The English language needs improvement.

5. The conclusion needs improvement. 

6. What is the advantage of this method over ANN, ANFIS,etc?

7. There are other refs related to maintenance that must be mentioned.

 

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 4 Report

The paper presents a multi-target regression algorithm based on multi-layer sparse structure for predicting the cost of ship equipment grade maintenance. The structure of the paper is clearly stated, including the presentation of contributions in the introduction, related works, the proposed approach, experiments, results, and conclusions.

However, the paper needs to be improved on the basis of the following comments and suggestions:

1. In the introduction, the authors write about a few methods dedicated to predicting the costs (lines 40-45): “In view of the complexity of cost driving factors and the difficulty of obtaining the cost variation knowledge with small data size, existing studies mainly use methods such as gray theory[2], case-based reasoning[3, 4], support vector regression[5-7] to model and predict in different field. However, the above methods are often based on a single output structure when predicting the costs, and the output results are single and do not take advantage of the correlation information between different subitem costs.”

 However, the authors have not highlighted the methods that could be successfully used in predicting the costs, and that provide multiple outputs, for example, artificial neural networks. I suggest adding at least one paragraph about using artificial neural networks to the considered task.

2. Table 3 does not have the reference in the text. The authors have mentioned “Table 4” in line 372, but in fact, there should be reference to Table 3. Please, insert all 23 input features in Table 3.

3. The authors have mentioned “BP neural network” in line 410. I think that abbreviation “BP” is related to the back-propagation learning algorithm but it requires the clarification in the text. Moreover, there is the lack of information about the learning parameters related to “BP neural network”, and a tool that was used to generate the results.

 

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Round 2

Reviewer 1 Report

I spent a lot of time thinking about the originality of this paper. It is about whether the learning techniques applied in many studies can affect the maintenance of ship equipment. As a result of thinking as a researcher who has experienced both actual ship operators and managers, the answer is no. The variables described here would be expected on land, but I have never seen a case where the predictions were correct in ship management. However, my thoughts have no scientific basis. On the other hand, your research may have a scientific basis. Of course, I still can't agree on variables. Therefore, I will pass the decision of this paper to the editor.

Author Response

Dear reviewers，

Thanks very much for taking the time to review this manuscript. I appreciate all your comments and suggestions! Please find my itemized responses below and my revisions/corrections in the re-submitted files.

The maintenance cost prediction technology studied in this paper is not for the operation or management of a shipbuilding corporation. Instead, it focuses on the military's perspective and serves the Navy to develop reasonable maintenance funding. The object of this paper is the naval warships, not the general civilian ships. The cost forecasting technique in this paper can provide the appropriate funding support for different fleet agencies to develop detailed budgets for the planned repairs of fleet warships each year so that the military funding managers can make rough estimates for the scheduled maintenance costs of individual warships.

As for the variables selected in this paper, it is based on the performance parameters of the warship itself and the operating conditions of the repairing factory. Since the cost of warship scheduled maintenance is highly correlated with its performance parameters, the operating condition/repair technology of the repairing factory also greatly affects the cost of warship scheduled maintenance. Therefore, the construction of the variable system for cost prediction from these two dimensions is feasible, which can also be demonstrated by the experimental results of the paper.

Reviewer 2 Report

Dear authors, thank you for improving your work.

Author Response

Thanks very much for taking the time to review this manuscript. I appreciate all your comments and suggestions!