A Multi-Objective Crow Search Algorithm for Influence Maximization in Social Networks

Round 1

Reviewer 1 Report

This paper is titled “A multi-objective crow search algorithm for influence maximization in social networks”. The authors proposed a multi-objective crow search algorithm (MOCSA) is proposed to optimize the problem with maximum influence spread and minimum cost, based on a redefined discrete evolutionary scheme. The overall quality of the paper is acceptable but still needs some improvements as follows

·        Author must review some latest research published in 2021 and 2022 in the related work section.

·        Author must compare the proposed work with the latest research published in 2021 and 2022.

·        The abbreviation should be mentioned carefully and once, not every time, full form alone with an abbreviation.

 

·        English language must be reviewed; there are many grammar errors that result in ambiguous and incomprehensible sentences

Author Response

Please see the attachment

Author Response File: Author Response.pdf

Reviewer 2 Report

1. Abstract - Please summarize the results obtained by the proposed method quantitatively. 

2. Literature Review - More related works should be covered in this section. More critical literature analysis is required to summarize the existing works.

3. Why Crow Search Algorithm (CSA) is chosen to solve the proposed optimization problem considered in this study instead of other more recently proposed algorithms such as Arithmetic Optimization Algorithm, Gorilla Troop Optimizer and etc.? Please justify the selection of optimization algorithm.

4. My major concern in this paper is that the authors did not explicitly mention how the multiple objective functions issues are being handled? How to update the solutions given the presence of conflicting objectives? How to identify the best performing solution? The descriptions provided for MOCSA is more applicable to solve single objective optimization problems. A major revamp is needed in this part to explain how the conflicting objectives in multi-objective problems are being solved. 

5. The three peer algorithms (i.e., MOPSO, MOBA and MOBHO) selected for performance comparisons are quite outdated. More recent works should be used for performance comparison to verify the strength of MOCSA. 

6. Please revise the y-axis caption in Figures 3 and 4 to make it more reflective to the intended objectives.

7. For Figure 3, the proposed MOCSA did not show dominant performances than those of MPSO, MOBA and MOBHO. The performance of MOCSA is often outperformed by MOPSO in most selected benchmark. Thus, it is not true to claim that MOCSA has the best performance in maximizing the Influence Speed.

8. Figure 5 - How to calculate the fitness value F(x)? It has not been mentioned throughout the manuscript. Furthermore, what is the purpose of comparing the fitness value?

9. The Pareto fronts produced by different algorithms should be compared. 

10. There are no quantitative results reported in this manuscript. 

11. Surprisingly, the classic MOPSO can perform comparably with the proposed MOCSA. It seems not reason to design MOCSA with more complicated search mechanisms. Please discuss this issue.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 3 Report

In this paper, the authors proposed a multi-objective crow search algorithm (MOCSA) to optimize the problem with maximum influence spread and minimum cost, based on a redefined discrete evolutionary scheme. The parameter setting based on dynamic control strategy and the random walk strategy based on black hole are adopted to improve the convergence efficiency of MOCSA. The experimental results indicate that the MOCSA can provide better performance compared with other algorithms.

It is an interesting paper which combines the existing techniques such as crow search algorithm, discrete encoding scheme, random walk on black hole, etc. The paper is well presented in general. However, there are some errors and concerns in the paper. Some parts are not very clearly presented.  

- The memory update method for the variable m^{j, iter} used in Equations (2) and (3) is missing.

- The notations used in Equations (1) and (5) are not consistent.

- It should explain the meaning of the variable x in Equations (4) and (6).

- Equation (8) is not correctly expressed. Please double check the equation in the original referenced paper.

- The algorithm uses random Walk based on black hole to speed up the convergence. The local exploitation ability is improved by allowing a certain probability of escape. How about use multiple black holes to balance the local and global exploration? Will it improve the performance of the algorithm?

- In Algorithm 2, the randomly initialize position vector m is defined but is not used.

- For the experimental study, I’m not sure it is appropriate to compare the algorithms based on different running time (since generally more running time an algorithm uses, the better result it produces). Instead, it should compare the algorithms based on the same running time (i.e. set all algorithms to the same running time, and then compare their results).

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 4 Report

Comments for author File: Comments.pdf

Author Response

Please see the attachment.

Round 2

Reviewer 2 Report

The authors have addressed all comments given in earlier review. No further comments.