Fractional Frequency Reuse Optimal SINR Threshold Selection Using NIR and ISODATA

REVIEWER 1

CORRECTIONS AND COMMENTS

1.

The paper addresses issues of the development of cellular networks and pursues to identify questions regarding inter-cell interference by the re-use of fractional frequency. As such it assists in determining the thresholds at which subscribers fall in the inner and outer regions, as the final goal is to establish throughputs and fairness within the network.

Well noted

2.

The abstract of the paper provides a good overview about the intentions of the research paper, and explains the main terms used in the paper, and explains which are the main focuses of interest. However, it would have been advisable to resume just as comprehensively some of the most important conclusions drawn.

Captured in Lines 20 - 22

3.

The introduction is comprehensive and details well the arguments for the selected approach. The theoretical background is sound and the literature review in the field is well covered. It provides a well-substantiated argument for the selection of the literature used in building up the reasoning of their research paper.

Well noted

4.

The literature review is well-selected and presented in a comprehensive and explanatory manner, and making visible how the methodology of the present paper was developed. The built example regarding parameters is sound.

Well noted

5.

The paper continues by presenting and addressing the main issues that assisted in constructing the research methodology and selecting the parameters providing the required framework, and illustrates accordingly also the used algorithm for performing thresholding and frequency partitioning.

Well noted

6.

The methodology selected for calculating the results is well-substantiated, as further presented in the obtained results, which are displayed in a clear, visible and understandable manner.

Well noted

7.

The discussion extends the presented results by putting into discussion some of the main findings, and offers the premises for continuing this type of research.

Well noted

8.

The conclusions are delivering insights into the issues approached by the paper, but it would have been better to provide a more structured framework for the conclusions, and address them in more detail.

Conclusion comprehensively captured as per the tracked changes. This has been based upon the results from the ECDF curves, Edge UE throughput, average UE throughput and Fairness bar graph.

9.

The paper is a valuable contribution, and is relevant with respect to the stated goal, However, some minor language revisions are advisable. The paper is recommended for publishing, after the suggested improvements regarding conclusions content and language revisions.

All English language revisions indicated as per the tracked changes in several instances e.g. Lines 31, 35, 40, 44, 57 etc.

REVIEWER 1

CORRECTIONS AND COMMENTS

1.

English needs to be checked /lines 34, 71, 115/

Revisions made as per the tracked changes

2.

Many abbreviations and parameters in the formulas are not explained /BS, PF, terms in (3), (4)/

All abbreviations and parameters explained as per lines 109, 110 etc

3.

The algorithms NIR and ISODATA /Table 2/ are equivalent, except for line 5, where NIR or ISODATA are called. It is not necessary to write both, but only once and in line 5 to indicate the choice of NIR or  ISODATA.

The algorithm has been revised as per the comment and ISODATA part removed and commented as per line 5 of the algorithm

4.

The results shown in Table 3 completely coincide with the histograms of FIG. 2 - 6. In this sense, histograms are redundant.

The histograms are still valuable as they capture the distribution of the subscribers as per their SINR values. This is not represented in tables 3 and 4

REVIEWER 2

CORRECTIONS AND COMMENTS

1.

I propose its publication with the conditions of eliminating the elements similar to the one published by the same authors in 2021: “Optimal SINR Threshold Selection In Fractional Frequency Reuse Using Otsu’s Method“, in: INTERNATIONAL JOURNAL OF SCIENTIFIC & TECHNOLOGY RESEARCH ,VOLUME 10, ISSUE 01, JANUARY 2021 ISSN 2277-8616,pp.186-192”, ( https://www.ijstr.org/final-print/jan2021/Optimal-Sinr-Threshold-Selection-In-Fractional-Frequency-Reuse-Using-Otsus-Method.pdf).

While both papers focus on improving the throughput and fairness performance of FFR schemes, they explore different approaches to set the SINR threshold. The other one focuses on Otsu’s method while this one considers NIR and ISODATA.

The depth of analysis of the results of this paper is more elaborate.

This paper includes a discussion of the ECDF curves of UE throughput which was not discussed in the other paper. Empirical Cumulative Distribution Curves (ECDFs) are used to measure the network performance. ECDF curves are a measure of the probability or percentage of the number of UEs achieve a particular throughput or less of the desired throughput chosen. This information can help in obtaining a minimum network performance level.

This paper analyses the average edge UE throughput as opposed to only peak edge UE throughput in the other paper. Average edge UE throughput gives a better indicator of network performance. The peak edge UE throughput only shows the highest throughput achieved by one edge UE.

Only the most crucial elements, imperative for creating a complete and cohesive flow and understanding of this paper are retained. These include the cellular network model and FFR layout, RF channel model, Fairness index, the Proportionally Fair LTE scheduler.

We believe that any further subtraction from this paper would make it incomplete, and incomprehensible.

We therefore feel this paper is sufficiently differentiated.

REVIEWER 3

CORRECTIONS AND COMMENTS

1.

The abstract delivers useful information about the intents and purposes of the paper, and is interesting as it manages to create a link between the logic in natural/medical sciences and the logic of computerized networks. It introduces some of the terms operationalized in the paper and the methodology employed, thus providing a good overview of the entire paper.

Noted.

2.

The introduction is comprehensive and details in a sound manner the correlation between organic cellular networks, and their analysis and uses in medicine and the cellular networks of interest for the present paper and for the good functioning of networks regarding UE speed, number of cells, antenna pattern, etc. The theoretical background is solid and the authors attempt to create a certain order in the ideas used as substantiation. Nevertheless, the approach should have been a bit more specific as to why references to medical uses are necessary and relevant in the context. Hence, though there are rich specialized literature references, it would have been a better option to have at least some better structured references that would help more readers to understand respective analogies leading to the subsequent methodological approach.

Noted. The paper focuses on setting an SINR threshold. The novelty of the paper is how the approaches that have been previously used to threshold images is used to set an SINR threshold. Instead of pixel values as inputs, we use SINR values as the inputs.

3.

The methodology was chosen according to the objectives and the collected data, and has the merit of addressing a large theoretical basis. The examples employed render transparent the approach and are well chosen regarding the objectives of the paper.

Noted.

4.

The step-by-step approach elaborates generously on how the analysis framework was constructed, and how the results were interpreted. The results are clearly differentiated and discussed based on the data used in simulations. The graphic and table representations, just as in the other chapters and sub-chapters of the paper support the results obtained. The analysis and explicit presentation of the results are convincing and well-supported.

Noted.

5.

The discussion of the results is also well written from the perspective of the scientific objectives pursued, and the references to the results obtained convincing. Nevertheless, one recommendation is necessary in this context: the absence of pertinent conclusions that would round-up the structure of the paper, by mentioning relevant findings and conclusion, along with some recommendations and stated intentions about developing further research in the field of the authors.

Conclusion and future works included.

6.

Another recommendation would be to review the paper regarding especially the grammar aspect and use of tenses in English, along with some other necessary language revisions.

Grammar and language revisions done.

7

In conclusion, the paper is interesting, relevant for research in the field and it is good for publishing. Nonetheless, we recommend strongly the addition of some conclusions as mentioned above, and English language revisions.

Language revisions done and Conclusion included.

REVIEWER 1

CORRECTIONS AND COMMENTS

1.

#1. Must be revised by native expert. It previously has been emphasized. Still many linguistic flaws, and vague statement s can easily be found. I totally got disappointed with such revision or editing. Simply see L8, L17-18, L28, … In L17-18, you have the full format and then again in L26 repeated while wondering in keywords you have used abbreviations???? What does it mean for used numbers in Keywords????? Many other problems also still in the context and thus cannot be assigned as any scientific edited version.  

#4. As mentioned in #1, this comment hasn’t been satisfied. Simply I asked what the reason is for using capital letter and you responded ‘…has been edited …’ while simply L8 in Abstract shows obvious conflicting with your response.

#5. Incredible response. You just have added the abbreviations in the keywords and then claim for whole changing??? Abbreviations must be mentioned in the first use not in keyword!!!!!

We have revised the paper for spelling and grammatical errors as per the tracked changes. The comments made on the abbreviations has been addressed. See Lines: 24, 25, 28, 30 & 34.

2.

#3. Was lost with this response. Look at the Abstract and context.

‘The focus of this paper is to propose techniques used in image processing’????

The focus of this paper is to improve throughput and fairness in FFR schemes by setting the best SINR threshold. The threshold is then periodically updated to always represent the current status of the cell. This introduces dynamism in the setting of the SINR threshold.

3.

#6. Is not just for citation. Could you please highlight where in the instruction for authors using bolded formulation for parameters have been insisted on?

We have removed the bolded formulae as can be seen in the latest revision. See Lines: 83, 84 & 102 etc. See as per the tracked changes.

4.

#7. You have the line numbers, exactly give the lines that have been modified.

Well noted

5.

#8. ‘Edited’??? just that??? How can I follow the new editions??? Looking in the whole of paper again??? You must and have to say for example, ‘the introduction has been edited and new sub-headers are categorized in lines … while the other sub-headers including … have moved under new section entitled …’

Wondering Introduction with obviously unusual format. Mixing the literature review, models, layout, …???? 7 sub-headers for Introduction?????? Incredible arrangement.

Part of the contents in the introduction were moved to form a section of Literature review. The literature review includes sub-headers. Refer to Lines: 80, 88, 106, 12, 134, 143, 158 & 202

6.

#9. Not satisfactory. I emphasized on literature review, and you claimed ‘edited in Introduction’ while simple look at reference list shows that nothing has been happened. So, your response is totally unacceptable and not lied on truth.

No new references because as shown above, the literature review was mixed up in the introduction.

7.

#11. Not satisfactory and unacceptable response. Give exact comparison with previous works and models and strictly highlight the main significant novelty and contribution. This is very well-known procedure in all field of science. Form the whole introduction it is totally dull what the problem is and what goal you are pursuing? What is the main limitation of previous works and which gap of them is going to be covered here? With what method and why? what motives for? What is the main novelty in this work rather than other ones? The main significant contributions?... You just have 18 references which shows very shallow review and with a simple search many advanced works than you can be found.

These are captured in the introduction. Check Line 37 for the gap.

The focus of this paper is to improve throughput and fairness in FFR schemes by setting the best SINR threshold. The threshold is then periodically updated to always represent the current status of the cell. This introduces dynamism in the setting of the SINR threshold. Lines: 18 & 19.

The methods used are global thresholding techniques. Lines: 14, 15, 16 & L47.

Contribution and motivation are captured in Lines: 75 – 78.

The novelty is in the adaptation of Image thresholding techniques in wireless networks. In image processing, pixels are used to determine threshold of foreground and background. In our case we input SINR values and developing a threshold for FFR to separate subscribers into inner region and outer region. Native Integral Ratio outperforms Iterative Self Organizing Data Analysis and static FFR

8.

#12. Not justified and informative conclusion. Must be reformulated to depict the bolded findings. This is not lecturing paper.

The conclusion is that NIR outperforms ISODATA and static FFR. The ECDF curves, bar graphs for throughput and fairness clearly articulate this. Percentages have been used to show by how much NIR outperforms the other two approaches.

9.

#13. Bar graphs for validation???????? Although ROC is widely used in ML or AI-based model, but it pretty works for SNR detectors by plotting the probability of detection versus the probability of false alarm for a given SNR. Concerning this matter strongly emphasize to read https://www.bis.org/publ/qtrpdf/r_qt1803v.htm just as an example. Concerning the other mentioned items in this comment, it is severely advised to read more papers.

#15. Totally unacceptable response.

What about on covering the lack of ability to be spatially invariant to the input data?

#16. You are working with data, aren’t you? Each measured data has involved uncertainty, hasn’t it? So, it is clear you must have an evaluation for uncertainty involved in the model.

#17. The comment is obvious and clear.

It is also beneficial to see the comparison top-down enrichment and lateral connections or consistency the backbone architecture for detection. Therefore, in such cases comparing the taken time become important.

#19. Obvious comment. Try to understand and analyze it.

Feature selection suffers from increasing overfitting risk when the number of observations is insufficient. It also requires significant computation time when the number of variables is large. Where did you solve this matter or discussed on?

Our study focuses on the reported SINR values to allow us set a threshold. We are interested in improving throughput and fairness. We are not focussed on how the detection is done and if there are errors. We opine that this can be the basis of an entirely separate study that can be published. For now this was not within our scope of interest.

10.

#14. You must also give the interpretation to readers not only to the reviewer.

You have a bunch of figures without any interpretation. What can be depicted from them?

Check the sub-header 4.1 in L242 to explain the histograms. Sub-header under Line 251 to discuss the ECDF curves. Lines: 255 and 256 covers the throughput and fairness bar graphs

12.

#18. This comment doesn’t align on ML or AI-based. I asked did you code it or not? if yes with what programming language? Then in model output you have some inputs that based on them you have output. This means that each input may have individual effect on the output. This is the main reason why you must guide the readers or do a sensitivity analysis. This is a very clear concept not only in the field of electrical and computer science but also all fields that are involved with nonlinear analysis.

The LTE System-Level simulator that we are using is written in MATLAB, which is a multi-purpose programming language and numeric computing environment developed by MathWorks.

We have used the simulator with our own functions and scripts that enable us to: implement NIR and ISODATA thresholding algorithms, setup the simulations described in the paper, clean and process data obtained from the simulations, and visualize it in the graphs and charts presented.

The input to our dynamic FFR system is the reported SINR values of the UEs in the simulated network. The factors that affect the SINR value are discussed in section 2.3 and include the transmitter power, small-scale fading gain, receiver noise factor, pathloss and the interference from neighbouring base-stations. The relationship between these factors and SINR is known exactly (equation 3), and so, it is not the focus of this paper to build and analyze a model that can predict SINR.

Our focus is using the reported SINR values, obtained through simulation, to set the best SINR threshold to achieve best fairness and edge throughput.