Designing a Hybrid Equipment-Failure Diagnosis Mechanism under Mixed-Type Data with Limited Failure Samples

Round 1

Reviewer 1 Report

To solve the problem of limited failure samples, the authors proposed to combine over-sampling technique SMOTE-NC and generateive adversarial network ctGAN to balance the data. Experimental results indicated that the proposed  combined method outperformed the methods of augmenting the data with SMOTE-NC or ctGAN alone.

Presentation about experiment was not clear. The authors shall clarify how the model was trained and tested. In particular, how was the dataset divided to the training set and the test set? How many and what samples were generated by SMOTE-NC and ctGAN, respectively, in each method?

There are several editing problems or unclear sentences. The authors shall check the text thoroughtly. Some of examples are pointed out below.

Line 164 mentioned that the CatBoost classifier was employed to generate additional data. The sentence is not clear or inaccurate. Isn't that the additional data were generated by SMOTE-NC and ctGAN?

Line 186 referred the metric as "Balance Accuracy". Shall it be "Balanced Accuracy" as appeared in the column title of Table 5? "balance accuracy" also appears in the text many times.

Line 189 mentioned F1 score. However, the F1 score did not appear in the other places such as the experimental results. Instead, Balanced Accuracy was used in the experimental results.

Line 197~199 are not clear. Product ID is of categorical data. What is the domain size? That is, how many distinct values of Prodcut ID in the dataset?

Line 235 indicated RNF is 1.2%, which shall be 0.2%.

Line 290 mentioned failure data accounted for 3.4%. However, the sum of those in Line 291 was 3.8%.

Terms were used inconsistently. For instance, "CTGAN", "ctGAN", and "catGAN" appeared in the text.

Author Response

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

Author Response File: Author Response.pdf

Reviewer 2 Report

This paper is interesting, and relevant and proposed a technique for predicting limited failure data with high accuracy. The paper demonstrated the depth of knowledge, scholarly written, presented and organized both technically and logically. The methods and materials were clearly outlined and the results adequately evaluated. Moreover, the references are sufficient for the paper.

Author Response

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

Reviewer 3 Report

The comments of this reviewer are listed as follow:

1) References should be numbered in order, for example, references 22 and 23 are given as the first reference on page 1, line 30. All references should be revised and brought in order.

2)The introduction section must include background information related to the main topic of this manuscript and also advantages and disadvantages of previous works should consider and then the authors should describe how they solve the problem and what is the advantages of their method. The authors may want to consider reorganizing the first section for a better flow.

3) Add a table for previous works (the first column can describe the method of each work) with advantages and disadvantages columns of each method and in the last column, authors can describe how their method solves the limitation.

4) The last paragraph of the introduction does not describe the contribution of the work very well. Although the idea is interesting but couldn’t solve properly in a scientific way. In my opinion, it should be more attractive if this part is revised adding contributions in point.

5) The rest of paper is not describe in the last paragraph of introduction section. Add a paragraph at the end of introduction that describe the rest sections of paper.

6) Section 2 (Materials and methods) is not consider in details, all equations are missing, describe this section with details.

7) Add the algorithms of you method step by step in section 2.

8) Table 1 and equations 1 to 3 need to be cited, Add references for Table 1 and equations 1 to 3.

9) The lack of a list of abbreviations used in the work and symbols applied in equations significantly hinders a proper understanding of the text.

10) In Tables 5,7,9,11,13 and 14 authors compared the proposed method with other methods, Add references for other methods or did the authors implement all of the methods?

11) Table 1 is not cited in the text, all tables should cite in the paper.

12) The description of section 3 is not sufficient. Illustrate and consider this section with more details.

13) Section 6 - Conclusion is more of a summary of the article than a systematic ordering of the research results. I suggest changing the structure of this section in such a way that the reader can read what, according to the authors, is the main research achievement and what practical conclusions result from this research.

14) A review of the English language should be performed. A revision of the whole document is necessary.

15) References 15 to 20 are not cited in the paper, double check all the references, cite all the references in the paper in order.

Author Response

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

Author Response File: Author Response.pdf

Reviewer 4 Report

The manuscript presents an interesting topic and it might have a significant impact on the intended research topic. All sections of this manuscript have been well organized and written. Particularly, the results and discussion sections have been presented in an obvious and understandable way for other researchers who are interested in this research area. I recommend it for publishing in its current form.  

Author Response

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

Round 2

Reviewer 1 Report

The authors have responded to my concerns. I have no furhter suggestions.

Reviewer 3 Report

The paper is revised well and it can be published.