Algorithms for Load Balancing in Next-Generation Mobile Networks: A Systematic Literature Review

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

Peer Review for an Article on Algorithms of Load Balancing in Next-Generation Mobile Networks: A Systematic Literature Review

The manuscript provides a systematic literature review of load balancing algorithms for next-generation mobile networks using automatic, semi-automatic, and hybrid methods according to the PRISMA approach. It thoroughly analyzes several selected publications, categorizing algorithms and highlighting research gaps in context-awareness, resource allocation, and artificial intelligence.

The objectives, techniques, and conclusions of the research are somewhat well summed up in the abstract. Although the way it is now presented is good, it would be better if it paid more attention on how special the classification is and how crucial load balance is for next-generation systems.

Suggestions

• The authors should emphasize the role of the categorization of the algorithms into automatic, semi-automatic, and hybrid in the abstract section.
• Could the authors add a few lines to discuss the significance of the study for 5G network optimization and what has been proposed to be done?

The introduction is quite clear and brief, presenting the problem of Load Balancing in mobile networks. While it clearly defines the need for the study, the authors could afford more parameters and clarity to present the research gap.

Suggestions

• Could the author give some general discussion of challenges in mobile networks for particular loads in the new next-generation networks?
• To justify the systematic literature review, could the author clearly state the research gap, especially the lack of comprehensive studies on AI-based load balancing algorithms?
• Could the author give some examples of real-world scenarios?

The methodology section clearly delineates the PRISMA model, presenting the search terms, the database, and the inclusion and exclusion criteria.

Suggestions

• Could the author provide a flow chart to elaborate on the presented PRISMA methodology (mentioned in Fig. 1), which can make the screening process easier.
• Could the author explain how the same or similar search strings were used or modified for each database, including the specifics of the syntax or other constraints of the used databases (like ACM, IEEE Xplore)?

• To learn more, explain why you chose 2014–2023, which might hide or eliminate basic work on older mobile networks like 4G.
• Use a decision tree or criterion table to explain how the researchers classified algorithms as fully automatic, semi-automated, or a mix.

The results section analyzes and classifies the 45 publications selected for this paper into load balancing, resource allocation, and context-aware systems. Table 3 and Fig 5. show the analysis and categorization of algorithms into semi-automated, automated, and hybrid.

Suggestions:

• Divide the results into sections according to the type of algorithm used (semi-automatic, automatic, or hybrid algorithms) to restore structural coherence.
• Could the author improve Fig. 5 (barplot of algorithm types) by including percentage information above the bars to illustrate algorithm type distribution clearly?

• The pie chart in Fig. 4 is not enough; add a table showing how load balancing and context-awareness are shared in the reviewed articles.
• Could the author explain with specific examples from the evaluated publications how each algorithm type, such as reinforcement learning [37] or fuzzy logic [10], solves load balancing problems?
• Apply the overall analysis to the performance indicators used in the studies to demonstrate the usefulness of these techniques.

 

 

The discussion section covers load balancing  emerging research areas of interest and new findings in high-speed mobile networks. It can also list the following regarding the scalability and limitations of the algorithms.

Suggestions

• Could the author provide a research question to clarify, how the systematic literature review findings differ from the prior literature and help explain the paper's conclusion?
• What are the author's predictions for algorithm development, including automation level and semi-automation?

 

Conclusion Suggestions

• It is not wrong to repeat the main findings in the conclusion, but the discussion of the insights should be narrower.
• Specify how the work contributed to developing context-aware and AI-based network optimization to catalyze future research.

 

Technical and Language Corrections:

• Ensure consistent use of terms like "load balancing," "resource allocation" and "context aware" across the paper.
• Simplify some of the more complex sentences to improve overall readability for a wider audience.
• Standardized the formatting of abbreviations like “ML” vs  “Machine learning” in text and tables.

Figures and Tables Suggestions

• Could the author further improve the pie chart representation in Fig.4?
• Could the author improve Fig. 5 by adding data labels?
• Could the author provide the comparison table contrasting the performance metrics like energy efficiency, latency, etc.
• The author should add a diagram that will illustrate the typical load balancing for 5G HetNet.

 

Additional Suggestions:

• List the drawbacks of examined algorithms, such as frequent calculations, high-quality data dependence, or non-stationary applicability.
• Include an analysis of the relative costs and benefits of adopting versus not adopting AI-based load balancing algorithms.
• Indicate how these results relate to 5G/6G networks standards as postulated by them and their implications for network policy.

Author Response

Our response in the attached file.

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

The paper is clear and well written. A detailed bibliometric survey is carried out, covering a large body of articles; the methodology and criteria are precisely defined and accurately described. A large portion of the paper is devoted to presentation of the contents of the articles cited in the references. However, the discussion and conclusions seem somewhat weak. In my opinion, further discussion on the basic methodologies used in the cited works, as well as their merits and possible drawbacks / weaknesses, would significantly enhance the paper's readability and usefulness as reference material for future studies.

Author Response

Our response in the attached file.

Author Response File: Author Response.pdf

Reviewer 3 Report

Comments and Suggestions for Authors

This work aims to conduct a comprehensive analysis and classification of the algorithms used in this area of ​​study to identify those suitable for load balancing optimization in next-generation mobile networks. The article is generally interesting, but needs to be improved in terms of literature review, innovation, and experimental design. I agree that this paper can be published after revision. Here are some comments:
1. The paper mentions the study of automatic, semi-automatic, and hybrid load balancing algorithms, but does not clarify the specific differences in the application scenarios of the algorithms. It is recommended to add a definition of the research scope in the introduction, clearly distinguish the characteristics and applicability of the algorithms in different scenarios, and highlight the research innovations.
2. Regarding the literature review of the article, some of the work on machine learning and deep learning analyzed is relatively old. It is recommended that the authors analyze some recent related work, such as Process monitoring for tower pumping units under variable operational conditions: From an integrated multitasking perspective, Multi-hop graph pooling adversarial network for cross-domain remaining useful life prediction: A distributed federated learning perspective
3. Although the next generation of mobile networks is mentioned in the article, the effect of the algorithm in practical applications is not deeply explored. It is recommended to add practical application case analysis, such as specific application scenarios in 5G networks, to explain how the algorithm solves problems in practice, and to enhance the correlation between innovation and practicality.
4. Some key concepts, such as "deep learning" and "reinforcement learning", are only briefly mentioned in the article, without in-depth analysis of their internal principles and connections. It is recommended to add relevant theoretical foundation chapters to elaborate on the connotation, development context and application mechanism of these concepts in load balancing.
5. The load balancing models involved in the article, such as regression heuristic models, are not clearly explained in terms of their mathematical models and derivation processes. It is recommended to supplement the specific mathematical expressions, parameter meanings and derivation steps of the model to help readers understand it in depth.
6. Although the article mentions the impact of the algorithm on network performance, it does not explore the reasons and mechanisms behind it. It is recommended to analyze the results in depth and explain the reasons for the performance differences of the algorithms in combination with the theoretical basis.
7. The innovation of the article is not obvious enough, and further discussion may be needed

Author Response

Our response in the attached file.

Author Response File: Author Response.pdf

Reviewer 4 Report

Comments and Suggestions for Authors

There are some issues with the paper:
The research background, motivation, and significance of the paper have not been expressed.
2. Load balancing is not reflected, please add a description of this part.
The mathematical model in the Objective section is not reflected.
4. Missing chapters for current work.
5. Increase experimental comparisons in load balancing and provide more convincing experimental data.
6. As a review paper, it is recommended to have no less than 100 references.

Author Response

Our response in the attached file.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

no further comments 

Author Response

We thank you for your valuable comments that have helped us to improve our manuscript.

Reviewer 2 Report

Comments and Suggestions for Authors

I think the modifications performed by the authors are adequate; hence, I would like to recommend publication.

Author Response

We thank you for your valuable comments that have helped us to improve our manuscript.

Reviewer 3 Report

Comments and Suggestions for Authors

The revised article still has many problems:
1. From the overall research content, although the automatic, semi-automatic and hybrid load balancing algorithms are classified and discussed, the classification method of this type of algorithm itself lacks novelty, and fails to propose an original algorithm classification system or conceptual framework from a new perspective. It is mediocre in terms of innovation and is only a certain degree of induction and arrangement based on existing research.
2. Although the outlook on the development trend of mobile network load balancing algorithms in the future is mentioned, such as the continued existence of full automation and semi-automation, and hybrid algorithms as a transition bridge, these prospects are mostly simple speculations on the evolution path of existing technologies, lacking forward-looking and subversive innovative views based on in-depth theoretical analysis or unique insights, and it is difficult to lead the frontier development direction of this field
3. For the five key development trajectories proposed, the content is relatively general and broad, lacking specific and detailed innovative indicators or measurement standards to support the forward-looking and uniqueness of these trajectories, making it difficult for readers to clearly grasp their innovative connotations and boundaries.
4. Although concepts such as deep learning and reinforcement learning are mentioned, there is a lack of in-depth analysis of their core algorithmic principles. For example, the neural network architecture in deep learning, the principle of activation function selection, and the specific mathematical derivation process of value function and strategy function in reinforcement learning are not explained in depth. This makes it difficult for readers to truly understand the essential operating mechanism of these algorithms in load balancing from a theoretical level, weakening the theoretical rigor of the research.
5. The mathematical models such as the regression heuristic model involved in the article are not clear and complete when presenting their mathematical expressions. The definitions of some key parameters are unclear, and the assumptions, scope of application, and the relationship and difference with other mathematical models of the model are not detailed. It makes it difficult for readers to accurately grasp the theoretical basis and applicability boundary of the model.
6. The article does not construct a clear and systematic experimental framework, lacks a clear definition of key elements such as experimental purpose, experimental hypothesis, and experimental variables, and it is difficult to scientifically verify the validity and reliability of the proposed views and conclusions based on experiments, which affects the credibility of the research results to a certain extent and cannot provide a strong experimental basis for practical applications.
7. The literature review was not well revised
8. There was a lack of rationality and sufficiency in the selection of experimental indicators. The selected indicators were difficult to fully cover the key dimensions of the load balancing algorithm performance. For example, the comprehensive impact of the algorithm on resource utilization, energy efficiency, etc. was not fully considered, making the experimental evaluation results unable to fully present the overall performance of the algorithm
9. When analyzing the algorithm performance results, the deep-seated reasons and mechanisms behind the data were not fully explored. For example, the explanation of the performance differences of different algorithms in specific scenarios only stayed at the surface description, and did not deeply analyze how the internal structure of the algorithm, parameter settings and other factors interacted with the scene characteristics to lead to performance differences

Author Response

Our response in the attached file

Author Response File: Author Response.docx

Reviewer 4 Report

Comments and Suggestions for Authors

The author has made careful revisions based on the comments of the reviewers, and there are no other comments, which can be accepted.

Author Response

We thank you for your valuable comments that have helped us to improve our manuscript.