Prediction and Classification of Phenol Contents in Cnidium officinale Makino Using a Stacking Ensemble Model in Climate Change Scenarios

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

The authors investigate the use of a stacking ensemble model to predict and classify the phenolic content in Cnidium officinale Makino under various climate change scenarios. They utilize environmental data, physiological response indicators, and physiological activity indicators collected from controlled cultivation environments. By applying data augmentation techniques to address data imbalance, they develop and evaluate a stacking ensemble model that incorporates multiple base models and a metamodel. The authors claim that their approach significantly improves prediction accuracy and classification performance for phenol content grades compared to individual machine learning models.

The topic is worthy of investigation given the increasing importance of understanding the impact of climate change on the production and quality of medicinal plants, which has been a growing focus in recent literature.

I have some questions/concerns that need to be addressed first.

·       How does the use of tabular variational autoencoders (TVAE) specifically address the data imbalance issue in this study? Can the authors provide more detailed steps on how TVAE generates the augmented data, including any constraints or loss functions applied during the process? Additionally, how was the effectiveness of the augmented data validated beyond similarity measurements?

·       What criteria were used to select the optimal hyperparameters for each model using Optuna? Were there any specific ranges or distributions defined for each hyperparameter during the optimization process? Furthermore, were any cross-validation techniques or additional validation sets used to ensure that the chosen hyperparameters generalize well to unseen data?

·       The manuscript mentions the use of Spearman’s correlation coefficient for selecting base models. Could the authors provide a more detailed explanation of how this selection process was carried out? Specifically, how were the initial set of candidate models chosen, and what thresholds or criteria were used to iteratively select and eliminate models based on their correlation coefficients? Additionally, how was the diversity of the base models ensured to prevent overfitting in the stacking ensemble model?

·       In the discussion, the authors mention that temperature and vapor pressure deficit (VPD) have higher feature importance in predicting total phenol contents under the SSP5-8.5 scenario. Can the authors elaborate on the specific statistical methods or techniques used to determine the feature importance? Additionally, how do these methods account for potential multicollinearity among the environmental variables, and what steps were taken to validate the robustness of the feature importance rankings?

·       I wonder what underlying environmental or physiological factors contribute to the observed differences in model performance, and how were these factors quantified and incorporated into the model analysis?

·       I also suggest that the authors reduce the number of figures! (by combining or only keeping the important figures).

·       Figures 2, 3, and 4àThese figures compare the quartile distributions of original and augmented data but could be enhanced with clearer labels and more detailed captions explaining the significance of the comparisons.

·       Figure 7à The flowchart of the proposed stacking ensemble model is helpful but could benefit from a more detailed step-by-step explanation in the caption.

·       Performance Tables (Tables 3, 4, 5, 6, 8, and 9): These tables provide comprehensive performance metrics but would be more effective if they included brief summaries or key findings in the captions to aid interpretation.

·       The manuscript is lengthy, with some sections being unnecessarily long. Condensing the introduction and focusing results and discussion more tightly can improve balance.

 

 

Comments on the Quality of English Language

 The language is generally adequate but could be improved for clarity and readability by simplifying complex sentences, ensuring smooth transitions, and polishing grammar and syntax.

Author Response

Thank you for all the valuable recommendations and insights. Responses to the reviewer's comments have been prepared and are attached in a separate file.

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

[1] In line 75, please provide the full name of the "CNN model".

[2] current flowchart does not effectively illustrate the proposed model. Consider using a different style to enhance clarity and presentation quality.

[3] The outlines in Figures 2-6 and Figures 8-16 could be removed to improve visual consistency and focus.

[4] For better comparison, please position Figure 13 on the same page.

[5] The color bars in Figures 12 and 13 do not clearly present the results. Consider revising them for better visual representation.

[6] Clearly articulate the contributions and limitations of this study, along with suggestions for future work, in the conclusion section.

[7] Further revisions are necessary to enhance the quality of the introduction, conclusion, and figures.

Author Response

Thank you for all the valuable recommendations and insights. Responses to the reviewer's comments have been prepared and are attached in a separate file.

Author Response File: Author Response.pdf