All reviews of published articles are made public. This includes manuscript files, peer review comments, author rebuttals and revised materials. Note: This was optional for articles submitted before 13 February 2023.
Peer reviewers are encouraged (but not required) to provide their names to the authors when submitting their peer review. If they agree to provide their name, then their personal profile page will reflect a public acknowledgment that they performed a review (even if the article is rejected). If the article is accepted, then reviewers who provided their name will be associated with the article itself.
I am happy to inform you that reviewers are satisfied with the revisions made by you. Therefore, I am provisionally accepting the manuscript for publication.
[# PeerJ Staff Note - this decision was reviewed and approved by Sebastian Ventura, a PeerJ Section Editor covering this Section #]
The author completed the revision process successfully.
The authors present details of implementation.
The author explained the convergence figure. The chosen learning rate and parameters seems to make sense.
The author made the necessary changes in the manuscript. It is acceptable.
It is fine
It is fine
It is fine
As per the reviewer comments, kindly revise the manuscript and resubmit.
[# PeerJ Staff Note: Please ensure that all review comments are addressed in a rebuttal letter and any edits or clarifications mentioned in the letter are also inserted into the revised manuscript where appropriate. It is a common mistake to address reviewer questions in the rebuttal letter but not in the revised manuscript. If a reviewer raised a question then your readers will probably have the same question so you should ensure that the manuscript can stand alone without the rebuttal letter. Directions on how to prepare a rebuttal letter can be found at: https://peerj.com/benefits/academic-rebuttal-letters/ #]
- In defining improvements, there are two sentences which are "A classifier with high performance has been designed, which finds solutions to problems in learning networks thanks to its two-layer learning architecture" and "Proposed model can detect 6 different apnea types as multi-class is a candidate system to be a good decision support system with its high performance and comprehensive analysis structure in the field of sleep disorders." The author should delete these sentences because he already mentioned in another sentence the same thing.
- The sentence "All the results obtained in section 4 are explained in sections" must be checked in Grammarly. There are the same mistakes also in the rest of the manuscript.
- The sentence "For the decision-making layer, tests were carried out with 11 different machine learning models, including Logistic regression (LR), Random forest (RF), decision tree (DT), Logistic Regression (LR), Random forest (RF), Decision Tree (DT), Gaussian Naive Bayes (GNB), Linear Discriminant Analysis (LDA), Ada Boost, Gradient boosting (GB), ExtraTree (ET), Extreme Gradient Boosting (XGBoost), Support Vector Classifier (SVC), K-Nearest Neighbor(KNN) and the machine learning classifier which is used as meta-learner to guarantee to obtain best results have been employed in this work" is wrong. Please correct it. Used same ML name twice etc.
- In Figure 4, the author gave learning curves. If he wants to get smoother curves, while the epoch number is increasing, the learning rate can be reduced.
- In Table 1, the author presented results in different metrics like accuracy, specificity, AUC, etc. There is a specific reason for it. Because providing the results under the same metric is important for a fair comparison.
- Authors proposed a 2-layered design for the detection of sleep disorder and apnea's events separately. Instead of doing classification separately, it can be done in a single module? What is benefit doing separately?
- In Figure 1, after apnea detection, classification is continuing. It must be explained. Classification is done from the beginning just for 3 classes (obstructive, central, mixed) or not?
- 19 features belonging to the PSG records must be given in detail. What are these features?
- How is done of the separation(percentage) of test and training data? Validation is used or not?
- To recover the imbalanced data, preprocessing was done. But the data is still imbalanced. The author is using weighting between classes?
- The model design was given as keras.model.summary(). To better understand, it can be given a better form. Also, author used deep learning model but he did not give any explanation about training. The chosen batch-size, epoch number, optimizer, learning rate, etc.
- The shared source code is containing only meta-learners. Also, deep-learning models must be shared.
- The proposed method was only conducted on the collected dataset. To see the effectiveness of the proposed method, it can also be conducted on other shared datasets.
1) The problem statement is not clearly mentioned in the abstract
2) The literature survey seems to be just a summary. A critical analysis on the existing literature is missing.
3) What is the source of your data?
4) How did you get the features? Are they available from your dataset (OR) Did you use any methods to extract the features?
5) I couldn't find any explanation for the deep learning methodologies given in your article. What are the implementation parameters used?
6) How did you get the results in Figure 6? Are they implemented using the same dataset?
7) What is meant by sleep apnea events detection in figure 8?
8) What is the new contribution/novelty in your work?
9) How do you say that your methods are free from overfitting?
already given
already given
already given
All text and materials provided via this peer-review history page are made available under a Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.