Social Recommendation Algorithm Based on Self-Supervised Hypergraph Attention

Round 1

Reviewer 1 Report

This paper proposed a social recommendation algorithm that incorporates graph embedding and higher-order mutual information maximization based on the consideration of social consistency; the proposed measure has the room to be improved before the acceptance of the manuscript.

 

 

1) The reason for choosing self-supervision by the authors should be explained. Couldn't their problem be solved with self-supervision? Explain the advantages and disadvantages of the selected method.

 

2) Due to the complexity of the article's structure and the use of different symbols for equations, the authors should create a table to introduce the symbols used.

 

3) Figure 2- The framework of HGATH 232 is unclear, and it is necessary to specify the sequence of steps and provide comments related to each phase on the image.

 

4) The AGG function in Equation 2 needs to be clarified; it needs to be stated why it is used, and the authors should fully explain it.

 

5 ) In 1.2.2, it should be determined if there is a significant connection between these two methods and what results will result from their concurrent use, why the authors use Hypergraph and Attention together, and also explain to them about the multiple paths and Attention's behavior.

 

6) The number of evaluation criteria in the list needs to be higher, and the authors should also evaluate the system's accuracy in addition to the errors.

 

7) The manuscript needs to have a section called "Discussion," where the proposed method is discussed from different viewpoints.

 

8.The following papers on the same topic should be cited and discussed:

1. A Hybrid Method for Recommendation Systems based on Tourism with an Evolutionary Algorithm and Topsis Model (2022)

2. Addressing the cold-start problem using data mining techniques and improving recommender systems by cuckoo algorithm: a case study of Facebook. Computing in Science & Engineering (2020)

 

3. An Effective Explainable Food Recommendation using Deep Image Clustering and Community detection (2022)

Author Response

First of all, we would like to thank our reviewers and the Editor for their valuable and constructive comments for this revision. We have revised the paper in the light of their useful suggestions and comments, which help to improve the quality of this manuscript. The main modifications are summarized as follows:

1. According to the reviewers' comments, we have added a section on how to analyze the HGATH model in terms of the accuracy of the system based on two evaluation metrics, recall and NDCG.
2. The detailed explanations to the self-supervision and AGG function are also given.
3. Five papers have been added to the revised version, so the references have been rearranged. See references for details.
4. According to the reviewers' comments, we added a section to discuss HGATH model.
5. That part about the relate works has been revised.
6. According to the reviewers' comments, we added a table to introduce the symbols used.
7. The expressions of this paper have been polished by a native English editor who has helped the author correct the relevant grammatical errors.

Our detailed responses follow. The text in italics is taken verbatim from the reviewers’ comments, and the roman blue text is our response. And all revised parts are marked in yellow in this paper.

    

Author Response File: Author Response.pdf

Reviewer 2 Report

The authors propose an interesting approach for social recommendation, I would like to see the related works summarised in a table where the authors compare them with their proposed approach. Also, I would like to see a discussion section to discuss an aggregated summary of the analysis, limitations, etc. 

Finally, authors can benefits from the following relevant state-of-the-art:

Guo, Zhiwei, et al. "Deep learning-embedded social internet of things for ambiguity-aware social recommendations." IEEE Transactions on Network Science and Engineering (2021). 

Abu-Salih, Bilal, et al. Social Big Data Analytics. Springer Singapore., 2021.

 

 

Author Response

Subject: Response to the reviewers and the Associate Editor )

Paper Title: Social Recommendation Algorithm based on Self-supervised Hypergraph Attention

Authors: Xu Xiangdong , Krzysztof Przystupa, Orest Kochan

 

First of all, we would like to thank our reviewers and the Editor for their valuable and constructive comments for this revision. We have revised the paper in the light of their useful suggestions and comments, which help to improve the quality of this manuscript. The main modifications are summarized as follows:

1. According to the reviewers' comments, we have added a section on how to analyze the HGATH model in terms of the accuracy of the system based on two evaluation metrics, recall and NDCG.
2. The detailed explanations to the self-supervision and AGG function are also given.
3. Five papers have been added to the revised version, so the references have been rearranged. See references for details.
4. According to the reviewers' comments, we added a section to discuss HGATH model.
5. That part about the relate works has been revised.
6. According to the reviewers' comments, we added a table to introduce the symbols used.
7. The expressions of this paper have been polished by a native English editor who has helped the author correct the relevant grammatical errors.

Our detailed responses follow. The text in italics is taken verbatim from the reviewers’ comments, and the roman blue text is our response. And all revised parts are marked in yellow in this paper.

    

    

Yours Sincerely,

Krzysztof Przystupa

College of Tongda, Nanjing University of Posts and Telecommunications

Author Response File: Author Response.pdf

Reviewer 3 Report

The paper focuses on a neural network-based recommender system applied to a social competition.

The topic is not particularly new, but the paper has some strengths and there are no obvious errors in its model. However, the paper has important weaknesses, namely:

- the organization of the manuscript is totally unacceptable with only the introduction (which includes introduction, related work, hypergraph and model), experiments and conclusions sections.

- there are grammatical errors and typos

- MAE and RMSE are not the best way to measure the effectiveness of a recommender system.

However, there is potential in this proposal and I suggest reorganizing the manuscript, providing a careful reading, and evaluating the experimental results with another set of measures.

Author Response

Subject: Response to the reviewers and the Associate Editor )

Paper Title: Social Recommendation Algorithm based on Self-supervised Hypergraph Attention

Authors: Xu Xiangdong , Krzysztof Przystupa, Orest Kochan

 

First of all, we would like to thank our reviewers and the Editor for their valuable and constructive comments for this revision. We have revised the paper in the light of their useful suggestions and comments, which help to improve the quality of this manuscript. The main modifications are summarized as follows:

1. According to the reviewers' comments, we have added a section on how to analyze the HGATH model in terms of the accuracy of the system based on two evaluation metrics, recall and NDCG.
2. The detailed explanations to the self-supervision and AGG function are also given.
3. Five papers have been added to the revised version, so the references have been rearranged. See references for details.
4. According to the reviewers' comments, we added a section to discuss HGATH model.
5. That part about the relate works has been revised.
6. According to the reviewers' comments, we added a table to introduce the symbols used.
7. The expressions of this paper have been polished by a native English editor who has helped the author correct the relevant grammatical errors.

Our detailed responses follow. The text in italics is taken verbatim from the reviewers’ comments, and the roman blue text is our response. And all revised parts are marked in yellow in this paper.

    

    

Yours Sincerely,

Krzysztof Przystupa

College of Tongda, Nanjing University of Posts and Telecommunications

Author Response File: Author Response.pdf

Round 2

Reviewer 3 Report

Most of my suggestions have been correctly addressed and the paper is qualitatively improved. However, I still suggest making acronyms explicit before their use and evaluating other approaches, e.g. doi: 10.1016/j.is.2018.07.002

Author Response

Dear colleague, we appreciate your contribution to our work! Thank you for helping us to improve the quality of our paper!

We revised our paper and made acronyms defined right after their first appearance in the text, e.g. “…how to use graph neural network (GNN) model…”, “…with respect to Mean Absolute Error (MAE), and Root Mean Square Error (RMSE)…”, etc. Now we believe each acronym is firmly explained prior introducing it in the text.

We considered the approach you recommend in our paper. Now it is reference 5. I addition, we found one interesting approach described in https://doi.org/10.3390/app10228223. We also added it to the citations. Now it is reference 2.