Fine-Tuning Fuzzy KNN Classifier Based on Uncertainty Membership for the Medical Diagnosis of Diabetes

Round 1

Reviewer 1 Report

Below are some of my comments and suggestions.

1. It is essential to make sure that the manuscript reads smoothly- this definitely helps the reader fully appreciate your research findings.

2. The numbering of Figure 1 on page 25 is incorrect.

3. The quality of the figures is low.

4. Standard expriments have not been performed. For example, in Table 10, the results of the proposed algorithm are given for only one evaluation method, while the results of competing algorithms are given for both 80/20 and 10-fold. There are similar drawbacks to other experiments.

Author Response

Dear Sir

I want to thank you for your time and efforts in considering and reviewing my manuscript ( applsci-1563763) entitled “Fine-Tuning Fuzzy KNN classifier based on uncertainty membership for medical diagnosis of diabetes”. Your constructive comments provided valuable insights to refine its contents and analysis. We worked hard to address the issues raised as best as possible. The manuscript has been rewritten to reflect comments and suggestions. A revised manuscript with red color indicating changes.  Please, find attached a point-by-point response to comments. I hope that you find our responses satisfactory, and the manuscript is now acceptable for publication.

Author Response File: Author Response.docx

Reviewer 2 Report

The paper is devoted to the development of a method for the classification of diabetes mellitus for pregnant women. Diabetes mellitus, a metabolic disease in which blood glucose levels rise over time, is one of the most common chronic diseases today. The relevance of the study is dictated by the importance of accurately predicting and classifying diabetes to reduce the severity of the disease and start treatment at an early stage. One limitation is diabetes datasets, which are limited and contain outliers and missing data. In addition, there is a trade-off between classification accuracy and operating law for detecting diabetic infection.

The article proposes a method for classifying diabetes for pregnant women using the Pima Indian diabetes dataset. The method consists of three stages:

1. The preprocessing stage includes culling outliers, imputation of missing values, a standardization process, and selection of attribute features that improve the quality of the dataset.

2. The classifier uses the fuzzy K-nearest neighbors and modifies the membership function to be based on uncertainty theory.

3. The grid search method is used to achieve the best values for tuning fuzzy K-nearest neighbors based on belonging to uncertainty since there are hyperparameters that affect the performance of the proposed classifier.

The proposed tuned fuzzy K-nearest neighbors based on the uncertainty classifier deal with the degree of confidence handle the membership function and the law of action and avoid miscategorization. The proposed algorithm performs better than other classifiers evaluated in this paper, including K-nearest neighbors, Fuzzy K-nearest neighbors, Naïve Bayes and Decision Tree. The results of different classifiers in an ensemble can significantly improve classification accuracy. Tuned fuzzy K-nearest neighbors have a time complexity of O(kn2d) and a spatial complexity of O(n2d). The tuned fuzzy K-nearest neighbors model has high performance and outperforms others in all tests in terms of accuracy, specificity, precision, and average AUC with values of 90.63, 85.00, 93.18, and 94.13, respectively. In addition, the results of the empirical analysis of tuned fuzzy K-nearest neighbors, fuzzy K-nearest neighbors, K-nearest neighbors, Naive Bayes, and Decision Tree demonstrate the superiority of tuned fuzzy K-nearest neighbors in terms of accuracy, reliability, and specificity.

Despite the satisfying results of the research, some shortcomings need to be corrected.

1. The aim of the research should be defined.
2. The state-of-art approaches should be separated from the methods and models proposed by the authors.
3. In my opinion, it is unnecessary to represent the distribution of patients by the histogram (Figure 3).
4. It should be justified why such size of data samples of female pregnant patients is enough for analysis.
5. The novelty of the research should be highlighted.
6. It is recommended to include obtained numerical results in the Conclusion section.
7. It is recommended to include researches diabetes analysis with the same dataset into the sources analysis part. E.g.: DOI: 10.1007/978-3-030-66717-7_13
8. Some text in Figure 1 is missed (bottom right corner).

In summarizing my comments I recommend that the manuscript be accepted after major revision.

Author Response

Dear Sir

I want to thank you for your time and efforts in considering and reviewing my manuscript ( applsci-1563763) entitled “Fine-Tuning Fuzzy KNN classifier based on uncertainty membership for medical diagnosis of diabetes”. Your constructive comments provided valuable insights to refine its contents and analysis. We worked hard to address the issues raised as best as possible. The manuscript has been rewritten to reflect comments and suggestions. A revised manuscript with red color indicating changes.  Please, find attached a point-by-point response to comments. I hope that you find our responses satisfactory, and the manuscript is now acceptable for publication.

Author Response File: Author Response.docx

Reviewer 3 Report

*) Please check formula (1). Perhaps the summation is not written correctly.

*) It would be important to give some indication on the computational complexity of Algorithm 1. This would allow us to understand if it is performing for any real-time applications.

*) In many formulations the standard has been used. I ask the authors if it has been verified whether the usual Euclidean space is the most suitable functional space.

*) Some captions are too poor. Please, if possible, make them self-explanatory.

*) The fuzzy classification techniques, as described by the authors in the paper, have now come into use in all those cases in which the data could be affected by uncertainties and / or inaccuracies. Lately, an interesting branch of fuzzy classifiers based on "fuzzy similarity" has developed. In other words, a data is classified according to its similarity (in the fuzzy sense) to a data class. Without prejudice to the fact that the authors' work (which I consider valuable) is beyond the application of these techniques, I still ask that a sentence be inserted in the text that highlights the possibility of using the potential of these techniques by putting the following relevant works in the bibliography:

doi: 10.1016/j.eswa.2008.09.015

doi: 10.1515/phys-2020-0159 

This last work is very interesting because it presents a fuzzy classification technique based on completely data independent fuzzy similarity that offers the possibility to group the doubtful cases in an additional class for any additional analyzes.

Author Response

Dear Sir

I want to thank you for your time and efforts in considering and reviewing my manuscript ( applsci-1563763) entitled “Fine-Tuning Fuzzy KNN classifier based on uncertainty membership for medical diagnosis of diabetes”. Your constructive comments provided valuable insights to refine its contents and analysis. We worked hard to address the issues raised as best as possible. The manuscript has been rewritten to reflect comments and suggestions. A revised manuscript with red color indicating changes.  Please, find attached a point-by-point response to comments. I hope that you find our responses satisfactory, and the manuscript is now acceptable for publication.

Author Response File: Author Response.docx

Round 2

Reviewer 1 Report

The article is acceptable in its current form.

Reviewer 2 Report

Thanks to the authors for consideration of reviewer comments. In my opinion, now paper can be published in current form.

Reviewer 3 Report

All suggestions have been implemented. So, the the paper deserves publication.