Classification Performance of Thresholding Methods in the Mahalanobis–Taguchi System

Round 1

Reviewer 1 Report

This article mainly introduces the benchmark research of threshold method in MTS, including types 1 and 2, ROC curve, Chebyshev theorem and Box-cox. The structure and presentation of this paper is very good, I believe it could potentially contribute to MTS works in the state of art. 

However, it lacks novelty. 

I would recommend the authors to comparing with some recent MTS-based publications like a PSO-based thresholding method. 

And it is better to add some results from experiments, thus further supporting this paper.

I would recommend to accept this manuscript.

Author Response

• However, it lacks novelty.

- We have highligted the novelty of the study in the Introduction section (Section 1) in the second last paragraph of the section and on the last sentence of the paragraph. We also iterated the novelty of the paper in the Conclusion section (Section 6) on the second sentence of the paragraph.

 

• I would recommend the authors to comparing with some recent MTS-based publications like PSO-based thresholding method.

-We really appreciate and very grateful with the recommendation to further study on PSO-based tresholding method in MTS since we did not aware of this. To the best of the authors’ knowledge, PSO-based studies in MTS were focused more significant feature extraction as well as feature selection methodology. We shall include this recommendation for future studies a moving forward.

 

• And it is better to add some results from the experiments, thus further supporting this paper.

- Yes, we can not agree more than this suggestion and several additional optimization results were added into the paper. However, due to page limitation, we only add 6 additional figures on the optimization results of which 3 are in the form of Bar graphs and the other 3 are in the form of SNR Plot charts of the optimization results. These results were taken from 3 datasets.  Out of 20 datasets, we selected 3 datasets namely the Wdbc, the Spambase and the Medical Diagnosis of Liver Disease datasets since these datasets showed higher number of variable reduction as compared to the rest.

Author Response File: Author Response.pdf

Reviewer 2 Report

1-The mathematical equations 19-20-21 part 3.5. Box–Cox Transformation, show flaws so please read again the description of the algorithm and improve it.

2-The inspiration of your work must be highlighted.

3-Please report the time complexity of the proposed algorithm.

4- More experimental figures results are welcome. It is my opinion as reviewer.

5- What are the advantages and disadvantages of previously mentioned methods and techniques?

Author Response

• The mathematical equations 19-20-21 part 3.5 Box-Cox Transformation, show flaws so please read again the decsription of the algorithm and improve it.
• The mathematical equations 19-20-21 were originally taken from the literature of Sachin Kumar, Tommy W. S. Chow, and Michael Pecht (2010) which is cited in this paper. We have double-checked them and so far, we followed exactly the description of the mathemical formulation. Except that we replace the term x with MD_i as the name of the mathematical function to relate to MTS studies. This formulation is also alligned with the same mathematical formulation in the studies conducted by Zhi Peng Chang, Yan Wen Li and Nazish Fatima (2019). We truly appreciate the recommendation by the reviewer to relook those equations and corrected them accordingly. Please refer to the updated Equation 19, 20 and 21 in the revise manuscript accordingly.

 

• The inspiration of your work must be highlighted.

-We have highligted the motivation statement of the study in the Introduction section (Section 1) in the second last paragraph of the section and on the second last as well as on the last sentences of the paragraph. We also iterated the motivation statement in the Conclusion section (Section 6) on the second sentence of the paragraph.

 

• Please report the time complexity of the proposed algorithms.

-It is the intention of this paper to examine the ‘effectiveness’ of all algorithms in performing the thresholding process of all data only. We did not measure the efficiency (time complexity) of the algorithms. However, as suggested by the reviewer, we will put this into consideration for future works to examine the time complexity of the algorithms as stated in the Conclusion section (section 6).

 

 

• More experimental figures results are welcome.
• Yes, we agree and it was done. However, due to page limitation, we only add 6 additional figures on the optimization results of which 3 are in the form of Bar graphs and the other 3 are in the form of SNR Plot charts of the optimization results. These results were taken from 3 datasets. Out of 20 datasets, we selected 3 datasets namely the Wdbc, the Spambase and the Medical Diagnosis of Liver Disease datasets since these datasets showed higher number of variable reduction as compared to the rest.  

 

• What are the advantageous and disadvantages of previously mentioned methods and techniques?

-To the best of our knowledge, based on the review studies that we have conducted, the advantageous as well as the disadvategous of those techniques were not mentioned. The reported studies only focused in demonstrating the use of those techniques. And thus, it has become the intention of this study to investigate them

Author Response File: Author Response.pdf