Using Kernel Density Estimation in Knowledge Distillation to Construct the Prediction Model for Bipolar Disorder Patients

Round 1

Reviewer 1 Report

In this study, a prediction model based on kernel density estimation is presented for bipolar disorder patients. Especially using machine learning algorithms in the medical field is a popular research topic. Although the study is interesting, it lacks many of the things an article should include. First of all, it is not clearly stated why such a study was carried out. The novelty of the study is not understood as there is no literature research. English also needs to be improved. Here are my other suggestions:

1. The abstract should give more information about the problem.

2. In the introduction, the research question is not clearly understood. Only information about patients with bipolar disorder is given. However, it is not clear why machine learning and KDE should be used and how there is a need in the literature. As such, the original value of the study is not specified.

3. There is no literature study on the subject. A critical literature research is expected from the authors.

4. A comparative analysis can be performed if more machine learning algorithms are added for the experiments.

5. What are the parameters of machine learning algorithms? Has a separate method been used for optimization? For example GridSearch vs.

6. Table 1 and Table 2 can be divided into three separate tables instead of sub-headings.

7. There doesn't seem to be a big difference in the results when using only X and X and Xpdf. The scientific significance of the results can be examined with a statistical test. For example, McNemar test

8. In conclusion, suggestions for future work should be presented.

Especially sentence structures should be corrected. In the introduction, there are many sentences that are difficult to understand and the context is unclear.

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 2 Report

Peer Review Report

Ms. Ref. No.: applsci-2558813

Title: Using kernel density estimation in knowledge distillation to construct the prediction model for bipolar disorder patients

 

Authors: Yu-Shiang Tseng, Meng-Han Yang

The subject of the article is within scope of the journal. The subject presented in the manuscript is very interesting and the results are promising. I recommend the paper for minor revision. Below you can find my main remarks about the manuscript.

Major comments:

1)      The introduction is poor. It should include much deeper literature review. The authors did not include information about already published paper in the area of application of machine learning techniques (used in this paper) for bipolar disorder prediction.

2)      The authors wrote only about the package, which was used for applying kernel density estimation. More information about the implementation should be given? What programming language? What for software was used to perform the calculations?

3)      The authors wrote: “Because the learning models in this study are all binary predictors of bipolar disorder …”. What it means? What are possible classes included in the database? It should be clarified in more details. Are there only 2 classes (healthy/disorder) or more?

Minor comments:

4)      Do not use and introduce abbreviations in the Abstract and Keywords of the manuscript. Simply use the whole terms.

 

Conclusion:

The subject is very interesting and the subject is within the scope of the journal. I recommend the paper for minor revision.  

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Round 2

Reviewer 1 Report

Thank you for the revisions. The work is acceptable in its current state.