Next Article in Journal
Water Reuse—Analysis of the Possibility of Using Reclaimed Water Depending on the Quality Class in the European Countries
Previous Article in Journal
Evaluation of the Level of Farmland Infrastructure Based on High-Resolution Images of UAV
 
 
Article
Peer-Review Record

Exploration of Biodegradable Substances Using Machine Learning Techniques

Sustainability 2023, 15(17), 12764; https://doi.org/10.3390/su151712764
by Alaa M. Elsayad 1,*, Medien Zeghid 1,2, Hassan Yousif Ahmed 1 and Khaled A. Elsayad 3
Reviewer 1: Anonymous
Reviewer 2:
Reviewer 3:
Reviewer 4: Anonymous
Reviewer 5: Anonymous
Sustainability 2023, 15(17), 12764; https://doi.org/10.3390/su151712764
Submission received: 2 July 2023 / Revised: 13 August 2023 / Accepted: 21 August 2023 / Published: 23 August 2023

Round 1

Reviewer 1 Report

The present study employed three different CARTs for the classification and feature selection of a biodegradation dataset. Please clearly state important findings you obtained in the present study while many studies have been conducted in machine learning.

Please provide an insight into the interpretability of your models using feature importance data as well as the accuracy, sensitivity, and specificity to further discuss the reliability and applicability of your models.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

In this manuscript, three different classification and regression trees (CARTs) were employed for the classification and feature selection of a biodegradation dataset. These models include the standard CART, CART with curvature test, and CART with curvature-interaction tests. Furthermore, to overcome the issue of overfitting and enhance model generalization, CART models were constructed using Bayesian optimization and repeated cross-validation. The experimental findings demonstrated that the proposed CART model, which integrates curvature-interaction tests, outperformed other models in classifying the test subset. It achieved an accuracy of 85.63%, sensitivity of 87.12%, specificity of 84.94%, and a highly comparable area under the ROC curve of 0.87. In the prediction process, the model identified the top ten most crucial descriptors, with the SpMaxB(m) and SpMin1_Bh(v) descriptors standing out as notably superior to the remaining descriptors. This paper is well written and meaningful, and the authors give adequate data to prove the conclusions. Therefore, I want to recommend it to be accepted after minor revision. Following are the comments:

1. Most figures are a bit blurred. Please change the figures with higher resolution.

2. The format of the references are not uniform. Please check carefully.

3. Please list the directory for the abbreviations shown in the manuscript.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 3 Report

Title not informative

The introduction could be merged with Literature Review, Additionally why the authors mentioned the aim of the work in the middle of the intro –this is confusing for readers please rewrite the intro in a new manner.

All figures need resolution enhancement as well as the permission status

Many abbreviations did not define and others were defined many times eg BO

Many eq were involved and some are unnecessary to mention in detail please check carefully

Why do the authors not mention the biodegradable materials as well as the blended nondegradable materials

Overall, the article needs a carefully serious revision

Some references could be helpful for authors

Immobilization of L-methionine γ-lyase on different cellulosic materials and its potential application in green-selective synthesis of volatile sulfur compounds

Ecofriendly bioactive film doped CuO nanoparticles based biopolymers and reinforced by enzymatically modified nanocellulose fibers for active packaging applications

 

 

 

 

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 4 Report

The authors evaluated a quantitative structure-activity relationship models based on classification and regression trees (CARTs) for biodegradability prediction. And then the obtained accuracies were relatively high. However, the discussion aspect is weak and needs to be improved. There is no comparison of the results with previous results or interpretation of how the results can be applied in practice. 

 

LL.94-96

Random Forests (RF) are an extension of the CART algorithm, utilizing the power of ensemble learning to improve prediction performance and reduce overfitting. And RF have been widely used to solve regression and classification problems.

Why didn't use the random forests?

 

3.2. Feature selection

Why did you only choose the three feature ranking techniques? Please clarify the reasons.

 

L.257

Reference is missing.

 

LL.405-406

Did you repeat the division procedure? 

 

Please increase the resolutions of the figures.

 

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 5 Report

The issue raised in the manuscript is quite important, it indicates which substances are readily biodegradable and also indicates the differences in selected substances. The time and method of biodegradation are also not indifferent to the environment.

The title of the work is consistent with the content. The summary is quite extensive, it could be shortened a bit.

The introduction gives a good background to the goal, analyzing the literature is thematically related to the work. The list also includes older items. The authors also cite their earlier works.

Graphs (proceeding algorithms) are an important element of the work and functioning of the analysis methods. They are very interesting, but it would be worth working on their quality, especially Fig. 2; 3; 4. The other figures also require minor corrections so that the graphs are clearer and more legible.

The adopted work methodology is correct and the methods are properly described.

Concluding conclusions are correctly formulated, but it would be worth highlighting them.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 3 Report

Accepted

Reviewer 4 Report

The authors added some elements to enrich manuscript.

Back to TopTop