Machine-Learning-Based Uplink Throughput Prediction from Physical Layer Measurements
Round 1
Reviewer 1 Report
In this work, the authors compare several machine learning algorithms for uplink data rate prediction. To this end, the LTE dataset is collected, and its parameters are determined accordingly. Experimental results show that Decision Treen and K-Nearest Neighbor algorithms perform well.
The novelty is limited. The work is nothing new but only evaluates the existing methods. The pipeline shown in Figure 4 describes a common way to train an ML algorithm. It is hard to recognize the contributions of this work.
In this case, this work is more like a report instead of a research paper.
Author Response
Dear reviewer,
First of all, thank you for your valuable comments. Let us clarify the novelty of our research. The previous studies have been focused on downlink channel prediction due to the much higher demand for downlink data rates. However, these studies and estimations have started to lose their significance due to the uplink traffic has skyrocketed with the wide-spreading uplink-dominant applications such as AV control, machine type communication devices (MTDs) smart body area networks (SmartBAN), IoT, wireless sensor networks, video conferencing, file sharing, VoIP, surveillance cameras, peer-to-peer (P2P) and cloud storage. Besides, there is very limited research on uplink throughput estimations and we also aimed to minimize the number of LTE physical parameters to increase the system robustness. Lastly, there is no study that compares uplink traffic prediction of three unique environments such as metropolitan, urban, and suburban using machine learning. The study reveals the best ML model for uplink throughput estimation if the environment type is provided.
Author Response File: 
Author Response.pdf
Reviewer 2 Report
Some suggestions for the study
- Include RQ and Hypothesis
- Evaluation at different confidence levels (on Hypothesis testing)
- Motivation is weak, please define and make it clear why this problem is important
- There are a lot of typos. Please check for typos and sentences (Example Abstract)
Author Response
Dear reviewer,
Thank you so much for your valuable comments. According to your comments, we have made the following changes:
Include RQ and Hypothesis: How well the machine learning algorithms will perform uplink throughput prediction for the specific environments such as metropolitan, urban, and suburban.
Evaluation at different confidence levels (on Hypothesis testing): In our research, we are not aiming to evaluate results at different confidence levels. In fact, we are targeting to optimize uplink throughput prediction by minimizing the RMSE values as indicated in Table 2.
Motivation is weak, please define and make it clear why this problem is important:
The problem has been defined and clarified as explained in the revised abstract and introduction and provided in the attachment. It is also explained here as well.
The previous studies have been focused on downlink channel prediction due to the much higher demand for downlink data rates. However, these studies and estimations have started to lose their significance due to the uplink traffic has skyrocketed with the wide-spreading uplink-dominant applications such as AV control, machine type communication devices (MTDs) smart body area networks (SmartBAN), IoT, wireless sensor networks, video conferencing, file sharing, VoIP, surveillance cameras, peer-to-peer (P2P) and cloud storage. Besides, there is very limited research on uplink throughput estimations and we also aimed to minimize the number of LTE physical parameters to increase the system robustness. Lastly, there is no study that compares uplink traffic prediction of three unique environments such as metropolitan, urban, and suburban using machine learning. The study reveals the best ML model for uplink throughput estimation if the environment type is provided.
There are a lot of typos. Please check for typos and sentences (Example Abstract): All the typos are double-checked and fixed.
Please, see the revised version of the research attached.
Author Response File: Author Response.pdf
Reviewer 3 Report
The authors have presented a well developed and proposed work along with all necessary calculations and details in brief.
Author Response
Dear reviewer,
Thank you so much for your valuable comments. You can also see the revised version of the research attached.
Author Response File: Author Response.pdf
Reviewer 4 Report
- Authors should further clarify and elaborate novelty in their introduction.
- What was the key motivation behind proposing the uplink data rate prediction?
- The introduction is not clear and very less literature is used. Follow these instruction: The introduction should briefly place the study in a broad context and highlight why it is important. It should define the purpose of the work and its significance, including specific hypotheses being tested. The current state of the research field should be reviewed carefully and key publications cited. Please highlight controversial and diverging hypotheses when necessary. Finally, briefly mention the main aim of the work and highlight the main conclusions. Keep the introduction comprehensible to scientists working outside the topic of the paper.
- Provide better quality image for Figure 10.
- Potential applications of the proposed method should be discussed.
- I would suggest authors to add a comparison with past studies, methods, and how your method is better.
- To have an unbiased view in the paper, there should be some discussions on the limitations of the proposed method.
Author Response
Dear reviewer,
Thank you so much for your valuable comments. Based on your recommendations the following changes are made.
Authors should further clarify and elaborate novelty in their introduction: Thank you for your remark, we have pointed out what is novel in our research paper in the revised version and summarized below.
There is very limited research on uplink throughput prediction, and we also aimed to minimize the number of LTE physical parameters to increase the system’s robustness. Besides, there is no study that compares uplink traffic prediction of three unique environments such as metropolitan, urban, and suburban using machine learning. The study reveals the best ML model for uplink throughput estimation if the environment type is provided.
What was the key motivation behind proposing the uplink data rate prediction?:
The previous studies have been focused on downlink channel prediction due to the much higher demand for downlink data rates. However, these studies and estimations have started to lose their significance due to the uplink traffic has skyrocketed with the wide-spreading uplink-dominant applications such as AV control, machine type communication devices (MTDs) smart body area networks (SmartBAN), IoT, wireless sensor networks, video conferencing, file sharing, VoIP, surveillance cameras, peer-to-peer (P2P) and cloud storage.
The introduction is not clear and very less literature is used. Follow these instruction:
The introduction should briefly place the study in a broad context and highlight why it is important. It should define the purpose of the work and its significance, including specific hypotheses being tested. The current state of the research field should be reviewed carefully and key publications cited. Please highlight controversial and diverging hypotheses when necessary. Finally, briefly mention the main aim of the work and highlight the main conclusions. Keep the introduction comprehensible to scientists working outside the topic of the paper.:
The required references (thirteen extra references) and literature review has been added to the introduction we also mention the main aim of the work briefly. Please check the revised version.
Provide better quality image for Figure 10.
The Figure 10 has been changed
Potential applications of the proposed method should be discussed.:
The potential applications of the proposed methods are included as follows
This application is very critical in terms of providing network services with adequate bandwidth where the environment has constraints such as shopping malls, concerts, fairs, and crowded areas needing a high uplink data rate. The proposed method will estimate the required uplink throughput and adjust the downlink-uplink channel bandwidth allocation asymmetry in favor of uplink. In other words, the allocated downlink channel bandwidth can be shifted to the uplink channel when needed.
I would suggest authors to add a comparison with past studies, methods, and how your method is better.
The Limitations & Discussion section has been added to the paper to discuss the comparison between the proposed methods and the other studies as well as the limitation of this endeavor.
To have an unbiased view in the paper, there should be some discussions on the limitations of the proposed method.
In the system, physical layer parameters must be obtained via receivers or scanners located in the experiment area to estimate the uplink throughput of a certain location However, receivers and scanners may not represent the same measurements as phones because of the differences in antenna gain. Crowdsource data measured and recorded with phone applications will also be sufficient for throughput estimation. Therefore, the proposed method should be also performed by mobile phones to see whether there is a significant change in measurements.
The Limitations & Discussion section has been added to the paper.
You can also see the revised version of the research attached
Author Response File: Author Response.pdf
Round 2
Reviewer 1 Report
My concerns are addressed.
Reviewer 4 Report
.
