REM-Based Indoor Localization with an Extra-Trees Regressor

Round 1

Reviewer 1 Report

This paper proposed a method for Indoor Localization based on REM and Extra Tree Regressor. The comments are as follows:

 1. Abstract: The abstract should emphasize the method introduced in this research, including the experimental setting, the hardware used for data collection, the model validation approach, and the evaluation metrics employed to assess research outcomes. Furthermore, a comparative analysis should be conducted to quantify the improvement achieved by this method compared to alternatives, highlighting its advantages.

2.  5. Machine Learning Regression Baseline Schemes: While the realm of machine learning encompasses a plethora of methods such as SVM, ANN, GPR, and deep understanding, this study exclusively adopts variants of decision tree-based approaches. This section should solely introduce the Random Forest Regressor, AdaBoost Regressor, and Decision Tree Regressor. A title revision for section 5 is recommended.

3. Figure 7: Although this study focuses on the Extra Tree Regressor, related diagrams and equations pertinent to this method are absent.

4. Addressing Comment 2: As suggested in comment 2, it is advised to augment the paper with a comparative analysis against other "State of the Art" (SOTA) methodologies to establish the superiority of the proposed approach.

5. Consideration of Training and Prediction Time: Training and prediction times are crucial in selecting machine learning methods. Including these time metrics as part of the method comparison evaluation is recommended

6. Measurement and Proposed Methodology: Sections 3 ("Measurement Methodology") and 4 ("Proposed Methodology") are suggested to be consolidated. This combined section should elucidate the research scenario, like area dimensions and grid size. Additionally, relevant details regarding the application of machine learning, such as dataset size, training and testing sample sizes, and model parameters, should be explained.

7. Table 2 and Figures: Please ensure that appropriate units of measurement are clearly labelled for both Table 2 and the figures.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

This article proposes an indoor positioning method based on Radio Environment Map. Overall, the article has a reasonable structure and provides sufficient technical details. The experimental results seem convincing enough. However, before being accepted, there are several issues that need to be addressed:

1) The abstract of the article is too long. The author spends too much space on discussing the research background. I believe the author should rewrite the abstract to better focus on the research contributions.

2) In Chapter 2, the author lists a lot of related works. To facilitate readers' understanding of the advantages and disadvantages of different positioning algorithms, the author should include a comparison table to compare the proposed algorithm with existing algorithms from different aspects.

3) In Chapter 1, the author mentions the concept of the Internet of Things (IoT). However, I believe this concept is not discussed sufficiently. In fact, indoor positioning is a typical scenario of the IoT or has evolved from existing IoT systems. The following works can provide some references for the author in terms of the IoT, Modeling and analyzing cascading failures for Internet of Things, Cascade Failures Analysis of Internet of Things under Global/Local Routing Mode.

4) In Chapter 6, I think more details of the experimental setup should be provided. Additionally, error bars should be included on the experimental curves.

5) One of the contributions of this article is improving the accuracy of the positioning algorithm using machine learning. The author should provide an analysis of the algorithm complexity and try to explain whether the complexity of the algorithm would affect real-time positioning.

The language quality is acceptable.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 3 Report

The paper proposes a methodology for constructing the Radio Environment Map (REM), which provides details on the radio frequency (RF) environment in a specific area. The REM is created by employing RF sensors to detect and examine signals from wireless devices like Wi-Fi networks and cell towers. The paper utilizes the extra trees regression (ETR) algorithm for indoor localization based on coverage prediction maps. The ETR scheme is optimized through parameter tuning using 10-fold cross-validation.

The reason for the size of the dataset and the limitations of the algorithm and regression should be justified, given the data acquisition process.

There is no scale label in Figure 3.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 4 Report

The paper is well written both grammatically and technically. Each part of the process is explained clearly and then results are analysed and compared, making it clear to anyone who is familiar with similar work. The work contained in this paper is novel and useful for the scientific community working in this field.

Discussions comparing and/or integrating with commercially available tools such as Ekahua, will be very valuable. 

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

No more questions!

Author Response

Thank you for your meaningful comments and valuable suggestions.