Memory-Based Sand Cat Swarm Optimization for Feature Selection in Medical Diagnosis

Round 1

Reviewer 1 Report

This manuscript does hard work in the experiment, however, there exist a lot of confusing parts in content.

1. Why the topic is focused on medical data? an excellent algorithm does not necessarily be on one area.

2. Abstract is too long for an academic paper.

3. In line 74, what is the memory-based strategy? is there any citation?

4. Between 89-90, where is section 5?

5. Section 2 is hard to read, it needs a concluded table to show the evolution of SCSO, authors, and their published year.

6. In line 169, what are the 3.0.1 means and so forth?

7. Algorithm 1 on page 7 reveals the basic sand cats swarm optimization

algorithm, however, where is the proposed algorithm?

8. In line 327, Binary Methods for Feature Selection is the one that was first proposed in academia?

9. In line 371, why 80/20 instead of 73/30 or cross-fold validation?

10. In line 381, where is the proposed BMSCSO?

11. In table 2, where is the setting of the proposed BMSCSO?

12. In line 419 and other lines, such as in line 587, where is Table 6.4 ?

13. The comparisons between the proposed BMSCSO and other state-of-the-art feature selection algorithms are found in Tables 3, 6.4, 5, 6, and 7. One can observe that BMSCSO is not always the best, then, what is the contribution of this article?

14.  In Table 10, BMSCSO is 4.095238 instead of 2.404761, which is not the best as said in line 593.

15. In line 628, the basic SCSO and proposed MSCSO are converted to binary format, and given the names BSCSO and BMSCSO, respectively, but, how to convert or where the algorithms are? 

The manuscript needs efforts to do revise to make it readable.

 

 

 

Author Response

We were very happy while correcting the errors you have stated explicitly. Indeed, we truly appreciate your valuable feedback and the time you spent in reviewing, commenting and correcting our work. We also admire your vigilance in finding and suggesting the correction for the errors. We have addressed all of your comments and errors as follows and we hope you find them satisfactory (The corrections are highlighted in blue).

 

Author Response File: Author Response.pdf

Reviewer 2 Report

The manuscript introduces two wrapper FS methods based on the sand cat swarm optimization algorithm for various medical datasets used in diagnosis.

Comparison results between the proposed BSCSO, BMSCSO and other methods was achieved using numerous metrics.

Here are some recommendations to enhance the quality of the study and manuscript:

1. Please highlight the contributions and the research question(s) in the introduction.

2. Only one fitness function is presented (in equation (21)). However, the authors said that "researchers working on multi-objective optimization problems could employ the proposed MSCSO". Please explain.

3. If Figure 1 is not yours, please add the source of the figure 1 (reference [39] I guess) 

4. Please justify the use of 80% of data for training and 20% for testing. This repartition is it based on empirical study or on recommendations from previous studies?

5. A complexity analysis of the proposed BSCSO and BMSCSO should be given to prove their applicability for large scale problems.

Author Response

We truly appreciate your valuable feedback and the time you spent in reviewing and commenting on our work. We also like your advice in correcting the introduction. We have addressed all of your comments as follows and we hope you find them satisfactory (The corrections are highlighted in orange) in the response file.

Author Response File: Author Response.pdf

Reviewer 3 Report

This manuscript presents an algorithm memory based sand cat swarm optimization (MSCSO) that can be applied for solving feature selection problems. MSCSO utilizes the exploration potential of SCSO to explore the search space quickly, and an internal memory component to converge more rapidly and identify the best global solution. To properly handle feature selection, the basic SCSO and proposed MSCSO are transformed into binary format and named BSCSO and BMSCSO, respectively. The algorithms, then tested on 21 medical datasets and the gained results compared with some comparative algorithms.

Overall, the manuscript presents another metaheuristic algorithm adapted for feature selection problem. However, the novelty of the proposed algorithm is not significant. The authors are encouraged to revise the current manuscript based on the following comments because some critical points have to be clarified or fixed.

 

1.      The Introduction section should be revised to clarify the research gap, explain the research methodology, and clearly state the aim of the study. Specifically, the authors should provide more information on the challenges of problems and how using metaheuristic algorithms addresses these challenges.

2.      The authors should summarize the study's main contributions in the introduction's final paragraph.

3.      The current literature review is shallow and requires a deeper review to support the study adequately. To address this, the authors should provide a detailed categorization of metaheuristic algorithms, including state-of-the-art and recently proposed algorithms, as well as their applications in solving real-world problems, such as an improved grey wolf optimizer for solving engineering problems, mtde: an effective multi-trial vector-based differential evolution algorithm and its applications for engineering design problems, dmfo-cd: a discrete moth-flame optimization algorithm for community detection, and ewoa-opf: effective whale optimization algorithm to solve optimal power flow problem.

4.      The use of transfer functions in this study demands further explanation. The authors should provide an overview of various techniques for binarization by considering the existing work "comparative analysis of transfer function-based binary metaheuristic algorithms for feature selection". This will enhance the readers' understanding of the methodology employed in the research.

5.      Regarding the wrapper feature selection methodology utilized in this study, the authors should refer to "a binary metaheuristic algorithm for wrapper feature selection" in line 99.

6.      It is worth noting that there are existing studies in the literature that specifically concentrate on medical datasets and disease diagnosis; such works include "binary sine cosine algorithms for feature selection from medical data" and "binary Aquila optimizer for selecting effective features from medical data: a covid-19 case study".

7.      The authors should provide context and explanations for subsections 4.1 and 4.2 to enhance readers' understanding of the study's methodology.

8.      The authors should specify which equations of SCSO are replaced by the proposed "Improved Initialization Process" and "Adaptive Update of Positions". This will assist readers in understanding the modifications made to the original algorithm.

9.      It is recommended that the authors include a pseudo-code of the proposed algorithm and a flowchart that highlights the changes made in MSCSO. This will aid readers in understanding the proposed modifications and their implementation in the algorithm.

10.   The authors should provide a sensitivity analysis for the parameter values of τ0 and τ1, as well as for a and b, as mentioned in line 375.

11.   The authors should provide an exploration/exploitation analysis to prove their following claim, "The key to BMSCSO’s acceptable level of performance is the algorithm’s sought-after balance between its exploration and exploitation features, which is made possible by the use of both the proposed mathematical model for this algorithm and the FS process."

12.   The authors should include a comparison of the convergence curve for the proposed MSCSO algorithm and comparative algorithms.

13.   The authors should

13.1.  clarify what "SCS" stands for in lines 332 and 333, "BSC" in 342, and "ISCSO" in line 372.

13.2.  use a consistent variable name throughout the manuscript, either D, dim, or j, to avoid confusion and ensure clarity.

13.3.  specify the equation number for the equation mentioned in line 374.

13.4.  review the manuscript and provide the correct references for the following sentence in line 376, "These values have frequently appeared in many related works [x, x, x, y]".

13.5.  correct the numbering of the table "Table 6.4" in lines 419, 422, and 485.

13.6.  provide the total number of datasets used for experiments accurately. They used 21 datasets. This should be corrected in lines 455 and 471.

14.   The authors should add references for

14.1.  QANA line 142.

14.2.  the UCI and KEEL datasets mentioned in lines 365 and 367. Also, provide references for the Leukemia, Colon, and Prostate_GE datasets mentioned in the manuscript to enable readers to access and verify the data used in the study.

14.3.  for the statistical tests, Friedman and Holm to enable readers to understand and verify the statistical analysis.

14.4.  The authors should clarify the reference number [77] mentioned in line 374 and ensure that it is consistent with the total number of references cited in the manuscript, which is 45.

Author Response

We really appreciate your valuable feedback and the time you spent in reviewing and commenting on our work. We have processed all your comments as follows and we hope you find them satisfactory (The corrections are highlighted in green) in the response file

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

I don't have further comments

Author Response

We would like to express our gratitude to you.

Reviewer 3 Report

Although the authors tried to address my comments, it is essential to note that some critical comments still require further consideration. To improve the quality of the manuscript, I would like to emphasize the importance of performing the modifications suggested in the previous round of review.

 1.      To address the authors' responses to each comment, it is imperative to submit point-to-point responses.

2.      The authors should submit the revised manuscript adhering to the specific format guidelines outlined by the journal.

3.      Comments #4 and #5 highlighted a critical gap in the manuscript's explanation of transfer functions and wrapper feature selection methodology. Further clarification on these aspects is necessary using "comparative analysis of transfer function-based binary metaheuristic algorithms for feature selection" and "a binary metaheuristic algorithm for wrapper feature selection."

4.      Regarding comment #8 and the lack of authors' point-to-point responses, the authors should specifically highlight which equations of SCSO are replaced by the proposed "Improved Initialization Process" and "Adaptive Positioning Update Process." 

5.      Comment #11 raised a significant concern regarding the authors' response to the exploration/exploitation analysis. Although some context and explanation were added in Subsection 6.5, it is critical to note that no results or curves were included in the revised manuscript. To address this gap, the authors must provide the necessary quantitative data, results, and visual representations (e.g., curves and graphs) to support the analysis of exploration and exploitation in the proposed methodology.

6.      Regarding comment #12, the authors should include a comparison of the convergence curve for the proposed MSCSO algorithm and comparative algorithms.

7.      Regarding comment #13.1, there is no clarification for "BSC."

8.      Regarding comment #13.5, there is no correction for the table numbering "Table 6.4".

9.      Regarding comment #13.6, the total number of datasets is not corrected in the revised manuscript.

Author Response

We would like to express our gratitude to you.

Please see the attached response file.

Best regards

 

Author Response File: Author Response.pdf

Round 3

Reviewer 3 Report

The authors have responded to most of the comments adequately, and I recommend the revised manuscript for publication.