Next Article in Journal
Short-Term Resilience of Soil Microbial Communities and Functions Following Severe Environmental Changes
Previous Article in Journal
Effects of Grass Silage Additive Type and Barley Grain Preservation Method on Rumen Fermentation, Microbial Community and Milk Production of Dairy Cows
 
 
Technical Note
Peer-Review Record

A Deep Learning-Based Model to Reduce Costs and Increase Productivity in the Case of Small Datasets: A Case Study in Cotton Cultivation

Agriculture 2022, 12(2), 267; https://doi.org/10.3390/agriculture12020267
by Mohammad Amin Amani 1,* and Francesco Marinello 2
Reviewer 1: Anonymous
Reviewer 2: Anonymous
Agriculture 2022, 12(2), 267; https://doi.org/10.3390/agriculture12020267
Submission received: 9 December 2021 / Revised: 6 February 2022 / Accepted: 10 February 2022 / Published: 14 February 2022
(This article belongs to the Section Artificial Intelligence and Digital Agriculture)

Round 1

Reviewer 1 Report

The manuscript addresses an original and interesting issue which is a case study using a deep learning approach on small data sets in cotton crop. However, the text needs further improvement, especially in language, methodology, and discussion of results. Therefore, I recommend Major Revision. The main considerations follow below:

 

General comments:

There are some spelling and grammatical errors throughout the text, so I recommend a careful review of the grammar and the English language.

In the methodology, the authors should provide more information about the computational resources used in the ML analyses, such as software, packages/libraries used, and the main parameters adopted to create the ML models, for example: batch size for all models; learning rate for the neural networks; cost parameter, kernel type (polynomial or) RBF and gamma value of kernel function for SVM;  maximum tree depth and whether pruning for decision trees; total number of trees for random forest. Or provide the references of the packages used if the models were run in the default package/software configuration.

The manuscript contains only the results without any discussion. A discussion of the results obtained, especially regarding the outperformance of the DNNs, and highlighting the practical implications of the obtained results for machine learning studies and specifically for the cotton crop need to be provided. Authors need to clarify the relevance, originality, and impacts of the results obtained based on the current related literature.

Below I list some references that could be included:

Osco et al. (2020). Leaf Nitrogen Concentration and Plant Height Prediction for Maize Using UAV-Based Multispectral Imagery and Machine Learning Techniques. Remote Sensing, 12, 3237.

Parent, L. E., Natale, W., & Brunetto, G. (2021). Machine Learning, Compositional and Fractal Models to Diagnose Soil Quality and Plant Nutrition. DOI: 10.5772/intechopen.98896

Specific comments:

line 12: replace “datsaets” with “datasets”

line 12: Remove “In particular…”

lines 33-35: provide citation

line 50: “Schuster et al was looking forward to finding…” provide the reference numbering

line 55: “Hong et al. examined 257 soil sample” provide the reference numbering

line 57: please give a definition of what extreme machine learning (EML) is.

lines 61-64: provide citation

line 87: replace “Hulugalle” with “by the authors” or similar

line 116: replace “Ph” with “pH” throughout the text

lines 135-136: please provide in the text or the caption the full name of the SMOTE technique mentioned in Fig. 1

lines 141-144: provide the meaning of the acronym SVC displayed in Fig.1

line 159: replace the term “Back-Propagation” with “backpropagation” throughout the text

line 200: replace “aren't” with “are not”.

lines 200-205: provide citation

line 211: replace “isn't” with “is not”.

lines 212-213: It is not clear what the Label-encoding technique is. Please explain better the Label-encoding procedure used

Author Response

We have addressed all comments indicated in the review report. All significant changes and modifications in response to the reviewer's comments are highlighted in the revised manuscript. To facilitate your review, in the revised manuscript, we have highlighted all significant changes. The yellow highlights show the answer to the reviewer's comments. Response to the same comment from two reviewers is highlighted in blue.

Additionally a thorough revision of language has been carried out to improve readability and understandability.

Please see the attached document.

Author Response File: Author Response.doc

Reviewer 2 Report

Comments

 

The authors propose a deep learning model as a viable approach to optimize information on the effect of a set of agricultural variables. In the study they analyze the soil with the aim of reducing cotton production costs.

The work is well structured, and the analysis performed through deep learning algorithms is correct.

 

The paper is relevant and therefore is a contribution to the area and future research.  Deep learning is still under exploration, so the results obtained show progress in this line of research.

 

There are only some aspects to improve, and that the authors should consider for a next revision of this paper.

 

Firstly, the authors indicate that soil analysis through deep learning will reduce production costs, since it is possible to combine the components of the soil and their quantities accurately. This should be explained in more depth, because such an analysis allows improving the quality of the product. The cost of production can be reduced when the logistics of the production process is improved. It is not clear which combination can reduce costs. There may be combinations that improve quality but do not reduce cost.

 

Second, the authors indicate that the objective of the paper is to discuss the application of artificial intelligence in agriculture, even in the common case of small data sets. This objective does not represent the contribution of the paper, as it discusses only soil data for cotton and not for agriculture in general. The authors should improve the wording of the objective, and clearly explain the contribution.

 

Finally, the data analysis is adequate, however, it is not clear how the uncertainty problem is solved. There is a lack of description of how this model will be applied in the future. For example, it may be through a decision support system, where data is updated every month, or with an IoT system. The analytics data expresses the results for a set of data on a specific set of variables. In the future this may change and the selection will need to be updated. This needs to be explained in more detail.

Author Response

First revision

We have addressed all comments indicated in the review report. All significant changes and modifications in response to the reviewer's comments are highlighted in the revised manuscript. To facilitate your review, in the revised manuscript, we have highlighted all significant changes. The green highlights show the answer to the reviewer's comments. Response to the same comment from two reviewers is highlighted in blue.

Additionally a thorough revision of language has been carried out to improve readability and understandability.

 

Please see the attached document.

Author Response File: Author Response.doc

Round 2

Reviewer 1 Report

The authors have made most of the requested corrections. However, the Discussion section still needs improvement. Therefore, I recommend Minor Revision of the manuscript. My comments are presented below:

The Discussion section should relate the findings to previous work in the field. Thus, a discussion based on results obtained in other related literature should be provided.  Currently there are several similar studies using deep learning (DL) approach for prediction and classification problems in various crops, and essential questions such as: what are the similarities and differences of this study to others that have used soil variables to predict cotton growth or yield? are there other studies using DL on small datasets? have these studies found satisfactory results? I believe that the Introduction has several passages that fit better in the Discussion, so the authors could relocate them in the Discussion section.

Author Response

Second revision

We have addressed all comments indicated in the review report. All significant changes and modifications in response to the reviewer's comments are highlighted in the revised manuscript. To facilitate your review, in the revised manuscript, we have highlighted all significant changes. The yellow highlights show the answer to the reviewer's comments.

Additionally a thorough revision of language has been carried out to improve readability and understandability.

 

 

REVIEWER 1

Please accept our deepest gratitude for reviewing our paper and recommending these valuable comments of correction. We hope this new version of the manuscript, which has been improved by your suggestions, proposes the research idea and the results better.

 

  1. The Discussion section should relate the findings to previous work in the field. Thus, a discussion based on results obtained in other related literature should be provided. Currently there are several similar studies using deep learning (DL) approach for prediction and classification problems in various crops, and essential questions such as: what are the similarities and differences of this study to others that have used soil variables to predict cotton growth or yield? are there other studies using DL on small datasets? have these studies found satisfactory results? I believe that the Introduction has several passages that fit better in the Discussion, so the authors could relocate them in the Discussion section.

Response: Dear respected reviewer, in order not to modify the readability and the completeness of the introduction we decided to keep the initial part as it was, while more explanations about the issues mentioned in the comment are added to the revised manuscript, in the discussion part (Yellow highlights – line 344-355).

We hope the referee agrees with our point of view. Conversely we are ready to move the paragraphs from the introduction to the discussion.

Author Response File: Author Response.doc

Back to TopTop