Predictive Modelling of Maize Yield Under Different Crop Density Using a Machine Learning Approach

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

The paper's overall structure is complete, the topic is relevant, and its use of machine learning to predict corn yield is particularly innovative.

 

However, the paper still has room for improvement:

 

1. The description in the article is rather confusing, and the logic needs to be clarified.

 

2. The paper mentions the use of a randomized complete block design (RCBD) design, but the description of the experimental randomization, replication, and field management (such as fertilization, irrigation, and pest control) is insufficient. It is recommended to supplement the experimental description of the randomized complete block design and detailed field management measures to enhance the reproducibility and reliability of the results.

 

3. The specific sampling times for the "four different time points" mentioned in the paper are vague. It is recommended to clarify the specific time points V1–V4. It is recommended to explain the reason for selecting these time points and their physiological significance during the corn growth period.

 

4. The paper mentions providing meteorological data charts, but these are not effectively incorporated into the prediction model. Meteorological factors (such as temperature and precipitation) have significant effects on yield and quality, and it is recommended that they be included as predictive variables in the model. 5. The article uses multiple machine learning models (XGBoost, Random Forest, linear regression, etc.), but the explanation of the model results is superficial and lacks in-depth analysis of key variables. We recommend adding SHAP value analysis or explanation of feature importance ranking, especially for XGBoost and Random Forest models.
5. The article uses cross-validation (CV), but does not provide an independent validation set to evaluate the model's generalization ability. We recommend splitting the training and test sets or employing time series cross-validation to more realistically evaluate the model's predictive ability.
6. The article primarily uses R² as an evaluation metric. We recommend adding metrics such as RMSE, MAE, and MAPE to more comprehensively evaluate model performance.
7. Some figures in the article have low resolution and unclear labels. We recommend improving the resolution and clarity of the figures and ensuring that all figures and labels and titles are legible.

 

Overall:

 

The topic of this article is relevant, the data collection is systematic, and the application of machine learning is innovative. However, due to insufficient description of experimental design, insufficient depth of model analysis, and inadequate interpretation of results, we hope that the authors will carefully revise the paper based on the above opinions to improve the scientific nature and readability of the paper.

Author Response

Dear Reviewer,

All comments are given in the uploaded file.

 

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

The manuscript presents a well-designed field experiment and integrates advanced statistical and machine learning approaches to predict maize yield under varying planting densities. The topic is highly relevant to current challenges in crop production and food security. The study is methodologically sound, and the results are clearly presented. However, I recommend minor revisions to address some specific issues noted below, and upon addressing these concerns, the manuscript will be suitable for publication.

Abstract: The abstract explains the study well. However, it is somewhat lengthy and repeats phrases such as “high accuracy and reliability.” It would be better to shorten sentences and focus on the main findings, such as the importance of moisture and the strong performance of hybrid H5. Ensure that accuracy values (R², percentages) are presented consistently.

Introduction: The introduction thoroughly covers maize density and yield, but some sentences are long and packed with references. Breaking these into shorter statements will make the section easier to follow. It would also be helpful to highlight the study's novelty in combining field data with machine learning.

Materials and Methods: The methods are detailed and transparent. However, the description of the field layout could be simplified. A brief table of soil properties and plot dimensions would be easier to read. In the statistical analysis subsection, the order of tests could be explained in simpler steps to guide readers who are less familiar with the procedures. Make sure all abbreviations are explained when first introduced in the text, not just in the abbreviation list.

Results: The results are well presented with tables and statistical outcomes, but the narrative sometimes repeats what is already shown in the tables. It would be better to highlight key patterns rather than re-listing p-values. PCA figures and their captions could also be shorter and more focused on the main insights.

Discussion: The discussion effectively links findings with previous studies. However, some background information from the introduction is repeated and could be trimmed. It would also be beneficial to briefly discuss why yield prediction models were less reliable and how future research could improve them.

Conclusions: The conclusions are clear but could be more impactful by emphasizing three key points: hybrid H5 performed best, moisture was the most reliable predictor, and machine learning shows promise with some limitations. Including a brief note on how this work could guide breeding or farm management would strengthen the closing message.

Author Response

Dear Reviewer,

Our response to your comments is given in the file uploaded.

Author Response File: Author Response.pdf

Reviewer 3 Report

Comments and Suggestions for Authors

The topic ‘Predictive Modeling of Maize Yield Under Different Crop 2 Density Using A Machine Learning Approach Title is an interesting topic which is timely. But it requires significant improvements, as detailed below

• Remove title from the topic, I think this is a mistake
• The order of writing a good abstract is not followed in this write-up. The objective is completely missing in the abstract. A good abstract should have the following in a logical sequence

1. Importance of the study
2. Objectives of the study
3. Methodology
4. Cogent results presentation
5. Conclusion

• Six hybrids cultivated at three planting densities were sub-15 jected to analysis of grain yield and yield components measured at four different time 16 points…… This statement is not clear. Which points ? on the field or time interval/scale ?
• Line 22: For two yield parameter ,…….. what do you mean by this ? what are these models, not stated
• Line 25 and 26: The models for moisture and oil content in maize seeds achieved 25 a high accuracy rate of 80% ….. what do authors mean accuracy? What parameters were used to evaluate it. Accuracy in modelling are evaluated using MAE, MBE, RMSE, NRMSE, and so on, while R-square value is for precision
• Line 41: Controlling plant density…. Modify this statement to, obtaining optimum plant density ….
• Line 106 – 108: Sow-ing of the trial 106 was carried out in three different crop densities, minimum, optimal and maximum 107 number of plants for the hybrids used in the trial (S1 - 40,816, S2 - 69,686, S3 - 98,522). ………. These planting densities have no unit. Please, include it. Why the choice of this range, justify
• The modelling part is poorly presented in the methodology. For example, what are the input and output variables? What percentage of the data was used for training, validation and testing.
• Moreso, background information on the machine learning algorithms used in this study are completely missing
• Model evaluation parameters are missing, for example MAE, MBE, RMSE and NRMSE. Their equations are expected in the methodology, but missing, and also their results are missing
• The results are incomplete without the inclusion of these evaluators for any modelling work. I suggest authors go back and compute all these and represent. Thank you

Comments on the Quality of English Language

Major English Editing Required

Author Response

Dear Reviewer,

Our response to your comments is given in the file uploaded.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

I am pleased to see that the author has made detailed revisions to some of the comments. I believe the current version is ready for publication.

Reviewer 3 Report

Comments and Suggestions for Authors

I have no further queries