Machine Learning Enabled 3D Body Measurement Estimation Using Hybrid Feature Selection and Bayesian Search

Round 1

Reviewer 1 Report

The article submitted for review describes a novel machine learning framework to predict the 3D body measurements. The authors proposed combination of the Bayesian search for fine-tuning hyperparameter and the feature selection for input data reduction approach that has not been presented in the literature so far. The authors comprehensively defined the current state of knowledge and described the proposed method in detail. The presented conclusions are supported by the results. The article is clearly written and has high scientific value. I believe it is ready for publication.

Author Response

We would like to thank the respected reviewer for the appreciation of our work
and the very positive feedback.

Reviewer 2 Report

In this paper, authors proposes an machine-learning-based approach for 3D body measurement estimation while considering the feature (measurement) importance. The feature importance is accounted for and the most critical features are selected to reduce the algorithm input features and to improve the performance of the ML method. The following review comments are recommended, and the authors are invited to explain and modify.

 

 Comment:  The abstract section is inconsistent and does not reflect the main contributions of the manuscript. The authors should rewrite the abstract section to mention the main purpose of the paper, primary contributions, experimental results, and global implications.

Comment: The design of networks and approach (Bayesian search for two machine learning methods RF, and XGBoost) are from my point of view outdate, nowadays completely replaceable by a deep architecture with modern modules or a fully connected transformer. Moreover, the new approaches work very effectively and quickly even in 3D.

Comment: “A preprocessing stage is applied to remove outliers and normalize the dataset”, how did authors apply that?

Comment: Nothing is mentioned about the implementation challenges.

Comment: The following clinical decision support systems using machine Learning, and medical imaging must be included to improve the quality of the paper.

·       10.1155/2022/2665283

·       10.3390/app9010069

Comment: Discuss the stability of the system in terms of complexity.

Comment: Could you please check your references carefully (in particular, proceedings: location of the conference, date of the conference, publisher's name and location...)? All references must be complete before the acceptance of a manuscript.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 3 Report

General comment:

Authors have ML tools for predicting 3D measurement gaps using cohort studies of different age group sample. Overall, the writing and analysis part is appealing. The results carry some merit. There can some discussion with respect to RF superiority from the published results.

 

Specific comments

 

L39: Instead of reference [13], mention it as Wuhrer and Shu [13]

Similarly, at L40, L41 and elsewhere in the manuscript the word reference can be substituted with the authors’ name(s)

L72: …… optimization is superior (instead of is the winner).

 L118: is it manual or using an algorithm? Elaborate

L120: How splitting sets were considered? Random or algorithmic approach?

Fig. 3 and Fig. 4 are generic representations of RF and XGBoost, respectively.

Table 2: continuous instead of continues                   

Table 4: Needs to include the computational platform features, eg: processor, RAM etc. in the materials and methods.

 

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 2 Report

The authors have answered my questions satisfactorily.