APGCN-CF: A Spatio-Temporal-Aware Graph Convolutional Network Framework for Minor Agricultural Product Recommendation in Rural E-Commerce

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

Using user similarity to recommend items is a well-known concept in collaborative filtering. User-based collaborative filtering approaches have long focused on detecting similar users (typically using cosine similarity or other metrics) and recommending items they have interacted with.
On its own, the concept of "top k similar users" followed by aggregating their interactions does not seem novel. It's a common idea in recommendation systems.
The approach is used on two minor agricultural datasets. However, it could have been applied to any other product dataset, whether agricultural or not. So, what is novel about applying this strategy to the two agricultural datasets?

 

Author Response

Dear reviewer,

Thank you for offering us an opportunity to improve the quality of our submitted manuscript (applsci-3477877). We appreciated very much your constructive and insightful comments. In this revision, we have addressed all of these comments. We hope the revised manuscript has now met the publication standard of this journal.

We highlighted all the revisions in yellow colour.

On the next pages, our point-to-point responses to the queries raised by you are listed.

Comment 1:

On its own, the concept of "top k similar users" followed by aggregating their interactions does not seem novel. It's a common idea in recommendation systems.

Response 1:

Thank you for your valuable suggestions. We fully agree with your observation that the concept of "Top k similar users" and aggregating their interaction behaviors is indeed a common approach in traditional collaborative filtering. We would like to clarify that the innovation of our proposed APGCN-CF framework does not lie in the mere use of user similarity concepts, but rather in the following aspects: First, our framework deeply integrates Graph Convolutional Networks with Collaborative Filtering, a combination that enables the model to simultaneously capture complex higher-order connection patterns and user-product interaction relationships. User similarity is just one component in our framework, which comprises four tightly integrated modules: graph structure construction, graph convolutional networks, candidate set generation (preliminary recommendation), and precise recommendation. Second, we innovatively leverage the characteristics of graph-structured data to incorporate temporal and geographical information through location-aware and time-sensitive embedding layers, enabling nodes to dynamically adapt to the characteristic variations of minor agricultural products. We believe this is crucial for recommendations of minor agricultural products that are sensitive to both temporal and spatial factors.

We have implemented the suggested revisions in our manuscript as follows:

1. In Section 1 (Introduction, lines 124-129), we have elaborated on the innovation points and contributions of our work.

2. In Section 4.4 (first paragraph, lines 495-511), we have emphasized that the innovation of our framework in finding similar users lies in utilizing the high-order feature vectors processed by graph convolutional networks. We have detailed the advantages of this method and restructured Section 4.4 to more clearly explain how the user high-order feature vectors obtained after graph convolution are used to find similar users. This makes the approach more comprehensible to readers and enhances the reproducibility of our work.

Comment 2:

The approach is used on two minor agricultural datasets. However, it could have been applied to any other product dataset, whether agricultural or not. So, what is novel about applying this strategy to the two agricultural datasets?

Response 2:

Thank you for raising this important question. We understand your concerns. Agricultural products have significant seasonal and regional characteristics compared to general commodities, which are even more pronounced in recommendations for seasoning-type minor agricultural products such as scallions and garlic. Unlike common agricultural products such as rice, wheat, and corn that are universally accepted and preferred, users' preferences for minor agricultural products are largely influenced by regional customs and dietary habits. Therefore, we aimed to incorporate geographical location and temporal information during graph structure construction, utilizing the powerful feature aggregation mechanism unique to graph convolutional networks to effectively capture user purchasing preferences that are difficult to identify using traditional methods. Experimental results demonstrate that when applied to minor agricultural product recommendations, our approach outperforms existing state-of-the-art methods in terms of precision, recall, and F1 score.

We have implemented the suggested revisions in our manuscript as follows:

1. In the Introduction section (lines 142-153), we have provided detailed supplementary information on why we chose scallions and garlic to validate our proposed framework. We emphasized the unique challenges in recommending minor agricultural products (such as scallions and garlic), clarified the differences between minor agricultural product recommendations and non-agricultural product or other agricultural product recommendations, further highlighting the novelty of our work.

2. In Section 2.3 (lines 288-309), we have added a comprehensive explanation of the unique challenges in the field of minor agricultural product recommendations. This helps readers clearly understand the challenges in recommending minor agricultural products before formally learning about our framework, enabling them to better comprehend the rationale behind our framework design and its distinctive features.

We have further emphasized these innovative points and unique contributions in the revised manuscript to make the value of the paper more clear. Thank you again for your valuable comments, which have been very helpful in improving the quality of our paper.

Your sincerely

Ke Zhu and Pingzeng Liu

Shandong Agricultural University, Daizong Street, Tai'an City, Shandong Province, China

On behalf of all the co-authors

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

Dear authors,

I have perused the article and the ensuing observations are appended below:

1. The utilisation of additional bibliographical references to substantiate section 2 is imperative, as they would assist in distinctly delineating the research's contribution with respect to extant publications.
2. The incorporation of further figures and tables in the exposition of the results is requisite, given the lack of clarity surrounding the results obtained from the research.
3. The methodology proposed is not clearly delineated, nor is the contribution of section 3.
4. The conclusions and discussion section is very limited, and it is recommended that the discussions be a subsection of section 5.

Author Response

Dear reviewer,

Thank you for offering us an opportunity to improve the quality of our submitted manuscript (applsci-3477877). We appreciated very much your constructive and insightful comments. In this revision, we have addressed all of these comments. We hope the revised manuscript has now met the publication standard of this journal.

We highlighted all the revisions in yellow colour.

On the next pages, our point-to-point responses to the queries raised by you are listed.

Comment 1:

The utilisation of additional bibliographical references to substantiate section 2 is imperative, as they would assist in distinctly delineating the research's contribution with respect to extant publications.

Response 1:

Thank you for your valuable suggestions. In the latest revised manuscript, we have incorporated the following additional references to support Section 2:

1. Bondevik JN, et al. A systematic review on food recommender systems. Expert Syst Appl. 2024 Mar;238:122166. (Line 298)

2. Song C, et al. Application of Intelligent Recommendation for Agricultural Information: A Systematic Literature Review. IEEE Access. 2021;9:153616-32. (Line 296)

3. Roy D, et al. A systematic review and research perspective on recommender systems. Journal of Big Data. 2022 Dec;9(1):1-36. (Line 212)

4. Ommane Y, et al. Machine Learning Based Recommender Systems for Crop Selection: A Systematic Literature Review. In: Machine Intelligence for Smart Applications: Opportunities and Risks. Cham: Springer Nature Switzerland; 2023. p. 21-59. (Line 317)

5. Aljunid MF, et al. A collaborative filtering recommender systems: Survey. Neurocomput. 2025 Feb;617:128718. (Line 218)

Comment 2:

The incorporation of further figures and tables in the exposition of the results is requisite, given the lack of clarity surrounding the results obtained from the research.

Response 2:

We greatly appreciate your valuable feedback on our paper. We acknowledge that the current figures in the results section were insufficient to fully and intuitively present our research findings. In the revised manuscript, we have added Figure 3 (page 17) and Figure 4 (page 19), which provide readers with a more intuitive comparison of the performance differences between APGCN-CF and baseline methods, as well as the benefits brought by incorporating graph convolutional networks for minor agricultural product recommendations.

Comment 3:

The methodology proposed is not clearly delineated, nor is the contribution of section 3.

Response 3:

Thank you for your valuable suggestions regarding the methodology section of our paper. We recognize that our description of the methodology in Section 3 was not sufficiently clear, and the contributions were not explicitly highlighted. In the revised manuscript, we have restructured Section 3 to more clearly describe how our method addresses the key challenges in minor agricultural product recommendations, while explicitly emphasizing both the theoretical and practical contributions of this section. Specifically, we made two major changes: (1) in the first paragraph of Section 3 (lines 327-337), we elaborated on the purpose and contributions of this section; and (2) in the last paragraph of Section 3 (lines 351-363), we supplemented the explanation of the significance and contributions of the methodology part, clarifying how this methodology helps design and validate the framework proposed in this paper.

Comment 4:

The conclusions and discussion section is very limited, and it is recommended that the discussions be a subsection of section 5.

Response 4:

We sincerely appreciate your suggestions regarding the conclusion and discussion sections of our paper. We fully accept your recommendation to include the discussion as a subsection of Section 5. In the latest revised manuscript, we have reorganized the structure of Sections 5 and 6, moving the original discussion from Section 6 to Section 5 as a subsection, and significantly expanded the discussion content to provide a more in-depth analysis and interpretation of the experimental results. We have also completely revised the content of Section 6, where we now clearly describe the work accomplished in this study and outline directions for future research. Specifically, the first change was to move the previous Section 6 discussion to become Section 5.7 (lines 804-834), where we have provided detailed explanations and discussions of the experimental results. The second change was to restructure Section 6 (lines 836-869) to clearly describe the work done in this paper and future directions.

We greatly appreciate your detailed feedback. These suggestions have been very helpful in improving the quality of our paper. We hope that this revision will meet with approval. Once again, thank you for your comments and suggestions.

Your sincerely

Ke Zhu and Pingzeng Liu

Shandong Agricultural University, Daizong Street, Tai'an City, Shandong Province, China

On behalf of all the co-authors

Author Response File: Author Response.pdf

Reviewer 3 Report

Comments and Suggestions for Authors

This paper aims to present a template/framework that aims to be innovative, and to a certain extent it succeeds, it is based on a convolutional network and aims to make recommendations for certain agricultural products, present in e-commerce in rural areas.

I consider the subject treated, and the way in which it is approached, to be relatively original, and quite relevant for the field addressed. The novel element proposed by the paper consists in the method of creating the framework intended for precise adaptation to minor scenarios of agricultural product recommendation, a method based on the construction of homogeneous and dimensional graphic structures, which are unified, to which is added the implementation of specialized modules for calculating similarity.

The structure of the paper is appropriate, the sections are balanced and follow logically, the way of presenting ideas is concise and uses appropriate technical language.

The technical way of writing the paper is in accordance with the journal's writing recommendations, and the quality of the figures and tables included is good.

The data set presented is correctly drawn up, but in my opinion, it needs to be expanded (I made recommendations in this regard to the authors), and the statistical analysis is based on a correct methodology that allows the verification and validation of the research hypotheses.

1. The analysis carried out by the authors, as a case study, takes into account only two agricultural products, namely: tea and garlic as minor agricultural products, which the authors consider representative and on which they focus their attention in the analysis of the models and the proposed method of their sales characteristics. In my opinion, in order to provide credibility to the proposed method and to demonstrate its validity, I believe that the analysis should be extended to other categories of agricultural products, taking into account seasonality and regional consumption levels. The study should include agricultural products available in different seasons, with different consumption levels, which are addressed to different categories of consumers and have different prices, all with the aim of providing a complete picture of how the recommendations made with the proposed method are made.
2. In subsection 5.2 the authors state that for the generation of personalized recommendations they use a mechanism that capitalizes on the similarity patterns of users. To make this mechanism easier to understand, I recommend that the authors provide more details about how similarity patterns are created/generated. Is this mechanism automated or can it be automated, based on a well-defined algorithm, to streamline the process?

The paper brings additional knowledge to the field addressed through the proposed method, and after making the indicated corrections, the scientific quality of the paper will be enriched, and the interest of the readers will increase.

The results of the analysis performed and of the entire research work are explicit and easy to understand for those familiar with the field, they represent a pertinent and valid scientific foundation for the conclusive part, which supports the objectives of the research carried out by the authors.

All these represent solid arguments that lead me to support the acceptance for publication of the paper, after making the minor corrections that I indicated to the authors.

Author Response

Dear reviewer,

Thank you for offering us an opportunity to improve the quality of our submitted manuscript (applsci-3477877). We appreciated very much your constructive and insightful comments. In this revision, we have addressed all of these comments. We hope the revised manuscript has now met the publication standard of this journal.

We highlighted all the revisions in yellow colour.

On the next pages, our point-to-point responses to the queries raised by you are listed.

Comment 1:

In my opinion, in order to provide credibility to the proposed method and to demonstrate its validity, I believe that the analysis should be extended to other categories of agricultural products, taking into account seasonality and regional consumption levels. The study should include agricultural products available in different seasons, with different consumption levels, which are addressed to different categories of consumers and have different prices, all with the aim of providing a complete picture of how the recommendations made with the proposed method are made.

Response 1:

We greatly appreciate your valuable suggestions for our research. We fully agree that expanding the agricultural product categories would enhance the comprehensiveness and universality of our study. Regarding our current choice of scallions and garlic as representative cases, we based our decision on the following considerations:

1. Scallions and garlic are the most fundamental seasoning-type minor agricultural products in Chinese culinary practices, with widespread market demand and stable consumption patterns, providing us with reliable benchmark data.

2. These two products have the most complete interaction records as well as user and product attributes in our accessible datasets, best supporting the thorough training and validation of our proposed model combining graph convolutional networks with collaborative filtering algorithms.

3. Scallions and garlic exhibit distinct seasonal and regional consumption patterns, which perfectly validates our model's ability to handle these unique characteristics of minor agricultural products.

We understand and value the importance of your suggestion to expand the analysis scope. In the revised manuscript, we have clearly indicated the importance of extending product category analysis in the "Future Work" portion of the Conclusion and Discussion section (Section 6), listing it as a significant direction for future work. Additionally, in Section 5, we have added an explanation for choosing scallions and garlic datasets for validation, emphasizing their unique characteristics as minor agricultural products with obvious seasonality and regionality. We plan to conduct more comprehensive experiments in subsequent research, covering various agricultural products with different seasonalities, price ranges, and targeting different consumer groups, to further validate and optimize our approach.The detailed explanation of dataset selection has been added in Section 5, lines 553-562 of the revised manuscript.

Comment 2:

In subsection 5.2 the authors state that for the generation of personalized recommendations they use a mechanism that capitalizes on the similarity patterns of users. To make this mechanism easier to understand, I recommend that the authors provide more details about how similarity patterns are created/generated. Is this mechanism automated or can it be automated, based on a well-defined algorithm, to streamline the process?

Response 2:

Thank you very much for your attention to the user similarity pattern generation mechanism in our framework. Yes, the generation of user similarity patterns in our proposed framework is defined by an explicitly automated algorithmic implementation. To make this mechanism more comprehensible to readers, we have reorganized and enhanced the explanation, improving the clarity and reproducibility of our paper. The detailed explanation of the user similarity pattern generation mechanism has been added in lines 509-527 of the revised manuscript.

We tried our best to improve the manuscript and we appreciate for your warm work earnestly, and hope that the correction will meet with approval. Once again, thank you very much for your comments and suggestions.

Your sincerely

Ke Zhu and Pingzeng Liu

Shandong Agricultural University, Daizong Street, Tai'an City, Shandong Province, China

On behalf of all the co-authors

Author Response File: Author Response.pdf

Reviewer 4 Report

Comments and Suggestions for Authors

The article has been revised to some extent. Yet, please take into consideration the following comments. 

1. You state that "rural e-commerce is important," but you can add more. For instance, provide statistics on the growth of this sector etc
2. Although the references are relevant, try to discuss them better than just listing the papers. Try to answer how they build towards the need for APGCN-CF.
3. Try to justify better the choice of sine-cosine encoding for time. Are there other options? Why do you prefer this method?
4.  Explain the rationale behind the used GCN architecture. Think about the layers, activation functions and hyperparameters.
5. Provide, if possible,e a stronger justification for the hybrid approach (USCG + PPPR)
6. Another direction would be towards analyzing more why GCN improves performance. I.e, discussing patterns that it can capture.
7. Discuss more the results and the limitations of the approach.

Author Response

Dear Reviewer,

We greatly appreciate the opportunity you provided to improve the quality of our submitted manuscript (applsci-3477877). We sincerely thank you for your constructive and insightful comments. In this revision, we have carefully addressed all these comments. We hope that the revised manuscript now meets the publication standards of the journal.

We have highlighted all revisions in yellow.

Below, we list our point-by-point responses to the issues you raised.

Comment 1: You state that "rural e-commerce is important," but you can add more. For instance, provide statistics on the growth of this sector etc.

Response 1: Thank you for your valuable suggestion. We have added a new paragraph in the introduction section (lines 33-46) that provides detailed data on China's agricultural product online retail sales in recent years. These data clearly demonstrate the rapid development and enormous potential of the rural e-commerce market, strongly supporting our argument about the "importance of rural e-commerce" and providing sufficient data support for the necessity of this research.

Comment 2: Although the references are relevant, try to discuss them better than just listing the papers. Try to answer how they build towards the need for APGCN-CF.

Response 2: Thank you again for your valuable insight. We indeed should discuss the role and value of references to our paper, rather than just listing them. To address this, we have thoroughly revised Section 2 (Related Work), making detailed adjustments in subsections 2.1, 2.2, and 2.3. The revised content no longer simply lists references but delves into the advantages and disadvantages of various studies, especially their limitations in handling minor agricultural product recommendations. We have also elaborated on how some key research inspired our work, building a clear logical bridge from existing research to the proposal of the APGCN-CF framework, and explicitly explaining why a new method is needed to address the special challenges in minor agricultural product recommendations.

Comment 3: Try to justify better the choice of sine-cosine encoding for time. Are there other options? Why do you prefer this method?

Response 3: Thank you for raising this important question. We have added detailed explanations in the time encoding section of Section 5.1 (lines 640-645), explaining the specific reasons for choosing sine-cosine encoding. We discussed the advantages of this encoding method in processing periodic time data, especially its applicability in capturing the seasonal characteristics of agricultural products, and compared the unique advantages of this method over others.

Comment 4: Explain the rationale behind the used GCN architecture. Think about the layers, activation functions and hyperparameters.

Response 4: We have supplemented the hyperparameter experimental results section in Section 5.6.3 (lines 924-933) with detailed explanations of the GCN architecture design choices. Specifically, we discussed the impact of different GCN layer numbers and hidden layer dimensions on recommendation performance, and analyzed these impacts from a theoretical perspective. We explained why specific layer and dimension configurations perform differently when processing datasets with different characteristics (such as garlic and scallion datasets), providing theoretical foundation and experimental support for our architecture design.

Comment 5: Provide, if possible, a stronger justification for the hybrid approach (USCG + PPPR).

Response 5: When introducing the framework in Section 4.1 (lines 482-490), we have added detailed explanations of the hybrid method (USCG+PPPR) design. We explained why we designed a two-stage recommendation framework and discussed how the hybrid method improves computational efficiency and reduces computational costs while ensuring recommendation quality.

Comment 6: Another direction would be towards analyzing more why GCN improves performance. I.e, discussing patterns that it can capture.

Response 6: We have conducted an in-depth analysis in the ablation experiment section 5.5.2 (lines 811-820), explaining in detail how GCN improves recommendation performance through its unique information transmission mechanism to capture high-order connection patterns between users and products. We discussed how GCN effectively aggregates neighbor node information, helping the model better learn user preferences and product characteristics, especially in sparse data situations. These analyses clearly demonstrate the key role and advantages of GCN in our framework.

Comment 7: Discuss more the results and the limitations of the approach.

Response 7: We have supplemented Section 6 (Conclusion and Discussion) with a more comprehensive and in-depth discussion of the method's limitations. We analyzed how computational complexity issues of our research method might limit performance on large-scale datasets, as well as deficiencies in the time information encoding method used, etc. We also discussed what methods we hope to adopt in future research to address these issues.

Thank you again for carefully reading our submitted manuscript and providing valuable suggestions, which have been very helpful in improving the quality of our paper.

Yours sincerely,

Ke Zhu and Pingzeng Liu

Shandong Agricultural University, Daizong Street, Tai'an City, Shandong Province, China

On behalf of all co-authors

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

The authors make some statements:

“First, our framework deeply integrates Graph Convolutional Networks with Collaborative Filtering, a combination that enables the model to simultaneously capture complex higher-order connection patterns and user-product interaction relationships. User similarity is just one component in our framework, which comprises four tightly integrated modules: graph structure construction, graph convolutional networks, candidate set generation (preliminary recommendation), and precise recommendation. Second, we innovatively leverage the characteristics of graph-structured data to incorporate temporal and geographical information through location-aware and time-sensitive embedding layers, enabling nodes to dynamically adapt to the characteristic variations of minor agricultural products. We believe this is crucial for recommendations of minor agricultural products that are sensitive to both temporal and spatial factors.

Unlike common agricultural products such as rice, wheat, and corn that are universally accepted and preferred, users' preferences for minor agricultural products are largely influenced by regional customs and dietary habits. Therefore, we aimed to incorporate geographical location and temporal information during graph structure construction, utilizing the powerful feature aggregation mechanism unique to graph convolutional networks to effectively capture user purchasing preferences that are difficult to identify using traditional methods. “

Still, no data or code is provided so it is impossible to evaluate at least at a high level what is actually used in the algorithm and if geographical location and temporal information has any role in improving the results. It is not even stated how the data was collected. Only a general description is provided.

 

Author Response

Dear Reviewer

Thank you for offering us an opportunity to improve the quality of our submitted manuscript (applsci-3477877). We appreciated very much your constructive and insightful comments. In this revision, we have addressed all of these comments. We hope the revised manuscript has now met the publication standard of this journal.

We highlighted all the revisions in yellow colour.

On the next pages, our point-to-point responses to the queries raised by you are listed.

Comment 1: Still, no data or code is provided

Response 1:

Thank you for raising these important issues. We understand your concerns and completely agree on the need for further work on research transparency and reproducibility. Regarding data sharing, we have submitted supplementary materials with our revised manuscript, including our preprocessed dataset, and we have added a data availability statement in line 967 of the main text. We have revised Section 5.1 (Experimental Datasets) in the main text, adding detailed information about the data sources, acquisition process, and preprocessing methods. We have provided a thorough description of the dataset contents, including new Tables 2, 3, 4, and 5 as dataset examples, along with detailed explanations of the meaning of each field. We have also specified the total number of products, users, and interaction records in our dataset and calculated data sparsity, aiming to help readers fully understand our dataset. Regarding code sharing, we have submitted the code for the main parts of this research with our revised manuscript. When running and testing, we ran the code in an environment with Python 3.12.4 and CUDA 12.1. Before running the code, please first configure the environment using the following command: "pip install pandas numpy transformers torch scikit-learn tqdm torch-geometric" and set the actual address of the data on your device in the corresponding parts of the code. We have provided detailed descriptions of our framework's operation in Section 4 (Proposed Work) to comprehensively help readers understand how our framework works.

Comment 2: it is impossible to evaluate at least at a high level what is actually used in the algorithm and if geographical location and temporal information has any role in improving the results.

Response 2:

Thank you for your valuable question. Demonstrating the role of temporal and geographical information in our framework is indeed essential. To address this, we have added Experiment 3 in Section 5.4 (specifically in lines 686-694), which is a new ablation study. We have validated the complete APGCN-CF framework against APGCN-CF-C (without geographical information), APGCN-CF-E (without temporal information), and APGCN-CF-F (without both geographical and temporal information) on both datasets.

The experimental results clearly demonstrate the significant contribution of temporal and geographical information to our framework's recommendation performance. In Section 5.5.3 of the revised manuscript, we have provided a detailed analysis of the experimental results(specifically in lines 765-800) and included Figure 5 to clearly illustrate the results of Experiment 3. We have also restructured Section 5.7 (Discussion) to include additional discussion of Experiment 3.

Comment 3: It is not even stated how the data was collected. Only a general description is provided.

Response 3:

Thank you for your valuable suggestion. As mentioned in Response 1, we have added detailed dataset information in Section 5.1 of the revised manuscript, including how the datasets were acquired and what information they contain.

In the latest revised manuscript, we have added more detailed dataset descriptions as per your guidance and included Experiment 3 to verify how the incorporation of geographic and temporal information improves the recommendation performance of our framework. Thank you again for your valuable comments, which have been very helpful in improving the quality of our paper.

Your sincerely

Ke Zhu and Pingzeng Liu

Shandong Agricultural University, Daizong Street, Tai'an City, Shandong Province, China

On behalf of all the co-authors

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

Dear authors:

Dear authors, thank you for considering the comments and making the corrections.

Author Response

Dear Reviewer,

We sincerely thank you for reviewing our revised manuscript and your positive feedback. Your professional feedback and constructive suggestions have played a very important role in improving the quality of our paper. We are truly grateful for the time and effort you have invested throughout the entire review process.

Yours sincerely,

Ke Zhu and Pingzeng Liu

Shandong Agricultural University, Daizong Street, Tai'an City, Shandong Province, China

On behalf of all co-authors

Author Response File: Author Response.pdf

Round 3

Reviewer 1 Report

Comments and Suggestions for Authors

I think enough progress has been made to warrent an accept recomandation.

Author Response

Dear Reviewer,

We sincerely thank you for reviewing our paper and recommending acceptance. Your professional feedback and constructive suggestions have played a very important role in improving the quality of our paper. We are truly grateful for the time and effort you have invested throughout the entire review process.

Yours sincerely,

Ke Zhu and Pingzeng Liu

Shandong Agricultural University, Daizong Street, Tai'an City, Shandong Province, China

On behalf of all co-authors

Author Response File: Author Response.pdf