Unsupervised Attribute Reduction Algorithms for Multiset-Valued Data Based on Uncertainty Measurement

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

The paper addresses the relatively unexplored problem of unsupervised attribute reduction in multiset-valued information systems (MSVIS), effectively bridging a gap in existing research, which has focused mainly on set-valued systems (SVIS). Introducing θ-tolerance classes and uncertainty measurements based on rough set theory and granular computing is novel and well-grounded. But there are several problems needs to be solved.

1. The related work section lacks recent developments in unsupervised feature selection, information granulation, and uncertainty modeling.
2. The choice of θ, which plays a crucial role in defining tolerance classes and information granules, is purely empirical. There is no automatic selection mechanism, nor is there a theoretical or heuristic guideline for optimal tuning. The selection of the parameter should be illustrated clearly.
3. The paper does not discuss the applicability of the proposed methods to high-dimensional, large-scale, or online learning scenarios.
4. The definitions and background theory, while rigorous, are overly detailed and repeated across sections.
5. The motivations of this paper is not clearly illustrated.
6. The section on theoretical preliminaries could be streamlined. Some parts—especially the mapping between multisets and distributions—may be moved to the appendix to maintain focus on the proposed method and results.
7. The English writing needs to be further improved.
8. Lack of new related references.

Author Response

Please see "Answer".

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

The paper is well written. Here are some minor comments :

1- The number of references is too large. I suggest removing the old one 

2-Some paragraphs are too large and some are small; it should be unified 

3- The style of the references is not the same 

4- The abstract is too large, it should be little bit shorter 

5- The paper needs some English improvements,

many grammatical mistakes see the discussion as an example, the last paragraph 

6- Algorithms 1 and 2 are very well Please add the codes in the appendix so all readers get benefits if possible 

7- Table 3. Details of the datasets for clustering 

Add some elaborations on the data used, how the data was collected, and why the authors chose these data?

8-Table 4. Optimal reduction results for k-modes clustering by the proposed algorithms 

This table is very important, but the table's size is large it should be fitted in the paper's margin

9-Figure 2. Clustering image of the original datasets with PCA

Add some elaborations in the  caption, what colors represent, and so on

10- Figure 7: The authors choose lambda=0.1, why?

What will happen if the value is changed?

Author Response

Please see "Answer".

Author Response File: Author Response.pdf

Reviewer 3 Report

Comments and Suggestions for Authors

The following must be amended/included:

1. The abstract must consist of a brief description of the results obtained.
2. The literature review must consist of more citations over the past three years.
3. Line 281, must provide a citation for UCI.
4. The conclusion must discuss how the results contrast or corroborate with the literature.
5. The conclusion must discuss the theoretical and practical implications of the research presented in the manuscript.

Author Response

Please see "Answer".

Author Response File: Author Response.pdf