Prediction of Strawberry Leaf Color Using RGB Mean Values Based on Soil Physicochemical Parameters Using Machine Learning Models

Round 1

Reviewer 1 Report

Title: multiple linear regression (MLR) and gradient boost regression (GBR) are really machine learning models.  I do not see the justification for mentioning mathematical models. The study does not have that scope. Remove that term from the title

Abstract: include RMSE metrics. The implication and relevance of the study is not clear.

Introduction

line 33: in recent decades?

line 40: in recent years?

line 44: .. impacted by growing conditions. why?

line 48: ..high-resolution cameras. what?

lines 95-117: the background on the selected models is insufficient, needs to be expanded and better justified.

lines 125-128: the scope and implication of the study is unclear.

 

Materials and methods

-the scope and implication of the study is unclear.

-a flow chart of the experimental and statistical methodology is required.

-Figure 4 lacks statistical significance. Furthermore, this figure and that subsection should be in results

-Figure 5. lacks components values. This figure and that subsection should be in results

-how you corrected for collinearity with the MLR model?

-What validation methods did you use for the models? How did you avoid overfitting?

 

Results

Performance training metrics ?

The results of the study are inconclusive, and the metrics are not the most optimal compared to deep learning for example.

Conclusions

conclusions should be reformulated

 

 

 

 

 

Author Response

Reviewer 1

Author response to the comments by the editor

Dear Reviewer

Journal - Agronomy

Manuscript ID: agronomy-1661210

Status: Major revision

I am Bolappa Gamage Kaushalya Madhavi. The first author of the above-mentioned article. That was reconsidered by the editor as a major revision. We are thankful to the editor and reviewers for carefully reviewing the manuscript and for the constructive suggestions offered. We greatly appreciate the valuable efforts of the reviewers to improve manuscript quality. A revision of the manuscript has been carried out with me and redeveloped the manuscript by addressing the comments given by reviewers. I believe that the revised version has been significantly improved and is suitable for consideration for publication in the MDPI Agronomy Journal.

We have done our best to address all of the points raised. Each comment is followed by the corresponding reply, which is highlighted in red colour. In addition, we made corrections and clarifications as the reviewers suggested in the revised manuscript.

Reviewer comments

Comments and Suggestions for Authors

Title: multiple linear regression (MLR) and gradient boost regression (GBR) are machine learning models.  I do not see the justification for mentioning mathematical models. The study does not have that scope. Remove that term from the title

Reply: Removed the term statistical from the title

Abstract: include RMSE metrics. The implication and relevance of the study are not clear.

Reply: RMSE metrics were included in the abstract along with R² values

Introduction

line 33: in recent decades?

Reply: Recent decades removed from line 38

line 40: in recent years?

Reply: Recent decades removed from line 50

line 44: .. impacted by growing conditions. why?

Reply: Strawberry leaf colour was changed according to the soil and outside environment. Therefore, replaced the conditions with the environment in line 46.

line 48: ..high-resolution cameras. what?

Reply: Added the examples for high-resolution cameras in line 51

lines 95-117: the background on the selected models is insufficient, needs to be expanded and is better justified.

Reply: Replaced one paragraph to lines 106 to 111 to make consistency and added a new paragraph in lines 126 to 129.

lines 125-128: the scope and implication of the study are unclear.

Reply: Removed some paragraphs and make the scope clear

 

Materials and methods

 

-the scope and implication of the study are unclear.

-a flow chart of the experimental and statistical methodology is required.

Reply: Added flowchart to make clearness in the procedure

-Figure 4 lacks statistical significance. Furthermore, this figure and that subsection should be in the results

Reply: It was replaced in the results session and added a paragraph in lines 312 to 327

-Figure 5. lacks components values. This figure and that subsection should be in the results

Reply: It was replaced in the results session

-how do you correct for collinearity with the MLR model?

Reply:  Pearson correlation heatmap was developed and checked the correlation coefficient values of every independent variable. According to the values of the correlation coefficient, independent variables are not highly correlated with another independent variable (not higher than 0.9). Therefore, every independent variable was selected for MLR and GBR models development.

-What validation methods did you use for the models? How did you avoid overfitting?

Reply: This experiment only performed greenhouse strawberry cultivation (the training 75% and testing 25%). Future, this research plan is to apply outside strawberry large scale farms for validation.

Here overfitting was avoided by dimensionality reduction by using “Principal Component Analysis (PCA)”. This analysis is a useful technique that can be used to mitigate overfitting in the machine learning model.

 

Results

 

Performance training metrics?

Reply: Training metrics were included in Table 2 and discussed in lines 404 to 415.

The results of the study are inconclusive, and the metrics are not the most optimal compared to deep learning for example.

Reply: Optimization was performed by hyper tuning the parameters namely estimators, learning rate, bagging seed, subsample, max number of leaves, max depth, max bin etc.

Conclusions

conclusions should be reformulated

Reply: The conclusion section added paragraphs lines 475 to 479 and 483 to 485.

Furthermore, scatter plots and histograms were added to improve the manuscript quality. Thanks a lot.

Best regards,

Kaushalya

Author Response File: Author Response.docx

Reviewer 2 Report

MLR and GBR models are used in this study to predict Strawberry leaf color from plant age and soil properties. it is a time series problem but regressions models are used for predictions. Time series model such as LSTM should better model the data. The manuscripts is well written but suggested changes and queries should be taken into account for its quality enhancement.

1. How the labelling is performed and did any expert provide the opinion?
2. RGB colors are sensitive to sunlight. How light exposure is controlled on strawberries?
3. Apart from color values, other measures such as entropy, energy and histogram values from leaf can be considered.
4. Texture of the leaf and its properties may also be considered for the prediction model.
5. A mathematical should be formed that would describe the relation of soil parameters with the color of leaf.
6. The intensity values of pixels and histogram of leaf on various days should be visually explored and plotted.
7. Last but not the least, a confusion matrix for the results and/or bar plots or line bars should be drawn to explain the results.  

Author Response

Reviewer 2

Author response to the comments by the editor

Dear Reviewer

Journal - Agronomy

Manuscript ID: agronomy-1661210

Status: Major revision

I am Bolappa Gamage Kaushalya Madhavi. The first author of the above-mentioned article. That was reconsidered by the editor as a major revision. We are thankful to the editor and reviewers for carefully reviewing the manuscript and for the constructive suggestions offered. We greatly appreciate the valuable efforts of the reviewers to improve manuscript quality. A revision of the manuscript has been carried out with me and redeveloped the manuscript by addressing the comments given by reviewers. I believe that the revised version has been significantly improved and is suitable for consideration for publication in the MDPI Agronomy Journal.

We have done our best to address all of the points raised. Each comment is followed by the corresponding reply, which is highlighted in red colour. In addition, we made corrections and clarifications as the reviewers suggested in the revised manuscript.

Reviewer comments

Comments and Suggestions for Authors

How the labelling is performed and did any expert provide the opinion?

Reply: Standard scaler was used to standardize data in the same unit variance and added the paragraph and mathematical equation in lines 210 to 220.

RGB colors are sensitive to sunlight. How light exposure is controlled on strawberries?

Reply: Frankly, this experiment only considers soil physicochemical parameters and plant age parameters for strawberry leaf colour change. It is a limitation of this research. However, the experiment was conducted in a controlled greenhouse for the parameters like temperature, and humidity. Future, we will plan to extend this research to sunlight and other environmental parameters' effects on leaf colour changes.

Apart from color values, other measures such as entropy, energy and histogram values from leaf can be considered.

Reply: This research only considers the R, G and B mean colour values for the strawberry leaf colour change. It is benchmark of this research. Future, it will expand to energy and histogram and entropy values comparison.

The texture of the leaf and its properties may also be considered for the prediction model.

Reply: This research only considered the colour changes. Because strawberry leaf colour is directly related with nutrition conditions in the soil. Because it is a shallow-rooted plant and soil nutrition is related to the leaf colour changes. According to that strawberry fruit quality and yield also reduced. Furture it can be spread out to texture analysis also.

A mathematical should be formed that would describe the relation of soil parameters with the colour of the leaf.

Reply: Mathematical summations were developed in the MLR model using regression coefficient values Eq. 7. 8 and 9.

The intensity values of pixels and histogram of leaf on various days should be visually explored and plotted.

Reply: It was added in Figure 5 for 60 days of leaf age  (vegetative stage) and 123 days of leaf age (reproductive stage)

Last but not the least, a confusion matrix for the results and/or bar plots or line bars should be drawn to explain the results.

Reply:  Scatter pots were developed for both MLR and GBR models using measured and predicted data in Figures 8 and 9.

 

Thanks a lot.

Best regards,

Kaushalya

Author Response File: Author Response.docx

Reviewer 3 Report

1. I am considering whether changes in leaf form and color due to seasonal or weather conditions should be taken into account.

2. There is a massive number of language issues present, requiring to correct every sentence.The authors should double-check the grammatical errors. And sentences are too complicated, meaning unclear, need to be simplified.

3. The picture inserted in the text is too blurry, need to replace it with a breath image.
4. Funding: Information regarding the funder and the funding number should be provided. Please check the  accuracy of funding data and any other information carefully.

5. The introduction is divided into 12 paragraphs, too many paragraphs, the content is cluttered, and needs to be reorganized. 

   Please refer to the current literature carefully, and make further supplement and improvement of the review part.

Author Response

Reviewer 3

Author response to the comments by the editor

Dear Reviewer

Journal - Agronomy

Manuscript ID: agronomy-1661210

Status: Major revision

I am Bolappa Gamage Kaushalya Madhavi. The first author of the above-mentioned article. That was reconsidered by the editor as a major revision. We are thankful to the editor and reviewers for carefully reviewing the manuscript and for the constructive suggestions offered. We greatly appreciate the valuable efforts of the reviewers to improve manuscript quality. A revision of the manuscript has been carried out with me and redeveloped the manuscript by addressing the comments given by reviewers. I believe that the revised version has been significantly improved and is suitable for consideration for publication in the MDPI Agronomy Journal.

We have done our best to address all of the points raised. Each comment is followed by the corresponding reply, which is highlighted in red colour. In addition, we made corrections and clarifications as the reviewers suggested in the revised manuscript.

Reviewer comments

Comments and Suggestions for Authors

I am considering whether changes in leaf form and color due to seasonal or weather conditions should be taken into account.

Reply: Frankly, this experiment only considers soil physicochemical parameters and plant age parameters for strawberry leaf colour change. It is a limitation of this research. However, the experiment was conducted in a controlled greenhouse for the parameters like temperature, and humidity. Future, we will plan to extend this research to environmental parameters / whether effects on leaf colour changes.

There is a massive number of language issues present, requiring correcting every sentence. The authors should double-check the grammatical errors. And sentences are too complicated, meaning unclear, need to be simplified.

Reply: Sentences again checked and corrected errors.

The picture inserted in the text is too blurry, need to replace it with a breath image.

Reply: Replaced with breath images

Funding: Information regarding the funder and the funding number should be provided. Please check the accuracy of funding data and any other information carefully.

Reply: Again corrected with adding funding number

Example article published this year from the same funding source in our laboratory. According to that again I corrected it. https://doi.org/10.3390/agriculture12020228.

 A Lightweight Attention-Based Convolutional Neural Networks for Tomato Leaf Disease Classification

 

The introduction is divided into 12 paragraphs, too many paragraphs, the content is cluttered, and needs to be reorganized.

Reply: Again introduction part was reorganised

Author Response File: Author Response.docx

Round 2

Reviewer 1 Report

Dear authors, most of the recommendations were correctly addressed, however, some additional clarifications and minor adjustments are required.

Materials and methods

- The validation method used to avoid overfitting was not properly justified. You only used the holdout method? why you did not use cross-validation or bootstrap sampling?, This, in order to improve the models performance.

- In the Statistical analysis subsection, what software, libraries or functions used for modelling?

Results and discussion

- In the Figure 9, you could add the adjustment metrics and significance

Conclusions

-Remove the "statistical" word

- The expression: "presumably" is inappropriate

- It is important to better conclude on the level of performance of these models in relation to plant age.

Author Response

Reviewer 1

Author response to the comments by the editor

Dear Reviewer

Journal - Agronomy

Manuscript ID: agronomy-1661210

Status: Minor revision

I am Bolappa Gamage Kaushalya Madhavi. The first author of the above-mentioned article. That was reconsidered by the editor as a minor revision. We are thankful to the editor and reviewers for carefully reviewing the manuscript and for the constructive suggestions offered. We greatly appreciate the valuable efforts of the reviewers to improve manuscript quality. A revision of the manuscript has been carried out with me and redeveloped the manuscript by addressing the comments given by reviewers. I believe that the revised version has been significantly improved and is suitable for consideration for publication in the MDPI Agronomy Journal.

We have done our best to address all of the points raised. Each comment is followed by the corresponding reply, which is highlighted in red colour. In addition, we made corrections and clarifications as the reviewers suggested in the revised manuscript.

Reviewer comments

Dear authors, most of the recommendations were correctly addressed, however, some additional clarifications and minor adjustments are required.

Materials and methods

- The validation method used to avoid overfitting was not properly justified. You only used the holdout method? why you did not use cross-validation or bootstrap sampling? This is in order to improve the model's performance.

Reply: First of all, thank you so much for your valuable suggestion. Frankly, in this research, we used only training and testing data only. We knew cross-validation or bootstrap sampling improves the model performance by a three-way holdout method (training, testing, and validation). Next research we plan to include cross-validation and implement seasonal changes that influence strawberry leaf colour changes.

- In the Statistical analysis subsection, what software, libraries or functions used for modelling?

Reply: It was included in the "2.1.4. Data preprocessing and models building"  section.

 

 

Results and discussion

- In Figure 9, you could add the adjustment metrics and significance

Reply: R­² values were added the above of all scatter plots

Conclusions

-Remove the "statistical" word

Reply:  Removed from the conclusion

- The expression: "presumably" is inappropriate

Reply:  Added "namely" instead of presumably

- It is important to better conclude on the level of performance of these models in relation to plant age.

Reply: Paragraph was added in conclusion part

 "Plant age also affected the skewed colour pattern. Furthermore, results indirectly revealed that increasing the plant age strawberry leaf R mean value was appreciably increased concerning G and B mean values, which leads to an increase in the model performance of R followed by G and B in both models"

 

Reviewer 2 Report

The manuscript is much improved and in an acceptable format. There are a few typos in the manuscript and thorough proofreading is required (such as line 190, using an image processing technique using an image segmentation algorithm and line 415 statical) etc.

In Equation 7, instead of constant value 107.58, a variable for example M1 should be introduced which may change by changing the data. Similarly variables in equation 8 and 9 may be introduced.  

Author Response

Reviewer 2

Author response to the comments by the editor

Dear Reviewer

Journal - Agronomy

Manuscript ID: agronomy-1661210

Status: Minor revision

I am Bolappa Gamage Kaushalya Madhavi. The first author of the above-mentioned article. That was reconsidered by the editor as a minor revision. We are thankful to the editor and reviewers for carefully reviewing the manuscript and for the constructive suggestions offered. We greatly appreciate the valuable efforts of the reviewers to improve manuscript quality. A revision of the manuscript has been carried out with me and redeveloped the manuscript by addressing the comments given by reviewers. I believe that the revised version has been significantly improved and is suitable for consideration for publication in the MDPI Agronomy Journal.

We have done our best to address all of the points raised. Each comment is followed by the corresponding reply, which is highlighted in red colour. In addition, we made corrections and clarifications as the reviewers suggested in the revised manuscript.

The manuscript is much improved and in an acceptable format. There are a few typos in the manuscript and thorough proofreading is required (such as line 190, using an image processing technique using an image segmentation algorithm, and line 415 statical), etc.

Reply: Removed "using an image processing technique" from the paragraph. Statistical also removed from the conclusion part

In Equation 7, instead of constant value 107.58, a variable for example M1 should be introduced which may change by changing the data. Similarly, variables in equations 8 and 9 may be introduced.  

Reply: Paragraph was added by introducing constant values

The constants of the MLR model are 107.58, 126.65, and 71.76 for R, G, and B mean values, orderly

 

Author Response File: Author Response.docx