RecPOID: POI Recommendation with Friendship Aware and Deep CNN
Round 1
Reviewer 1 Report
Comments for Authors:
Dear Authors,
After thoroughly reviewing your manuscript, I found it interesting. However, there are several issues that have to be addressed.
First, extensive editing of English language and style required. Without this, the paper is hard to follow. Further:
Abstract:
- Informative, brief and sufficient.
INTRODUCTION:
- If not necessary, please move the Figure to the next section.
- Contributions of the paper should be in the Discussion or Conclusion sections. Instead, a knowledge gap in literature should be addressed when it comes to your study.
MATERIALS AND METHODS:
- subtitles should be after the section title and no text should be between them. Lines from 117 to 120 should be under 2.1. as a short paragraph.
RESULTS:
- Figures should be mentioned/noted not far from the position of the Figure within the manuscript. A little restructuring is required so the explanations could be closer to the Figures.
- A DISCUSSION SECTION WOULD benefit your paper. The results section would be shorter, and the obtained findings could be described in more detail in this new DISCUSSION section.
CONCLUSION:
Limitations of your study should be addressed
Guidelines and recommendations should added.
Overall, the paper has potential, but first it requires major revisions when it comes to grammar and style and the overall structure of the paper.
Kind regards,
Reviewer
Author Response
RecPOID: POI Recommendation with Friendship aware and Deep CNN
Journal of Future Internet
Reviewer #1:
- Extensive editing of English language and style required.
RESPONSE: We are grateful to you for your insightful and helpful comments. We have revised the paper according to the comments. We tried to edit the article's English language grammar and style; we hope it will be accepted by you.
- If not necessary, please move the Figure to the next section.
Contributions of the paper should be in the Discussion or Conclusion sections. Instead, a knowledge gap in literature should be addressed when it comes to your study.
RESPONSE: Thank you for pointing to these items. According to your guidance, Figure 1 is deleted and the Contributions were moved.
The knowledge gap is added to the introduction as follows:
Three items are significant to POI recommendation: friend importance, check-in correlation, and user preference. Previous research reports one of these factors or does not adequately describe them.
However, most recent studies in location-based recommendation systems consider all social friends, although the survey of check-in pattern's most similar friendship is important to accurate prediction and recommendation according to the user's interests.
However, few studies have been proposed on location-based recommender systems with the aid of artificial intelligence and deep learning that incorporate key factors such as exploring the patterns of check-in behaviors and the impact of most similar friendships and user preferences.
- Subtitles should be after the section title and no text should be between them.
RESPONSE: I'm glad you showed me this. It has been removed.
- Figures should be mentioned/noted not far from the position of the Figure within the manuscript.
A DISCUSSION SECTION WOULD benefit your paper. The results section would be shorter, and the obtained findings could be described in more detail in this new DISCUSSION section.
RESPONSE: Due to your very good guidance, the Results section structure (subsection 3.1) has been changed, and the unit of Discussion on Performance (subsection 3.2) has been added.
- Limitations of your study should be addressed
Guidelines and recommendations should added.
RESPONSE: Based on your valued and effective points, the following content is added to the Conclusions:
This research's limitations can be the amount of data required for network learning, and also because clustering is used, there may be limitations within clusters. For example, in a region, we want to suggest several restaurants. If the restaurants are too close, the algorithm may mistake and put them all in one cluster. In future work, we plan to work on other datasets as well. We also intend to Consider reinforcement learning instead of convolutional networking because it has the advantage of learning with less data.
Finally, we would like to thank you again for the excellent comments and suggestions
for improving the paper.
Reviewer 2 Report
The paper discusses a new POI recommendation algorithm called RecPOI. RecPOI uses fuzzy clustering and CNNs for better prediction. recPOI takes into account both social and spatial context. The algorithm has been tested on well-known datasets and have been compared to a few other POI recommendation algorithms. Overall, the algorithm is novel enough to warrant publication. But, the paper needs comprehensive English language editing and better formatting. Shorter sentences will help structure statements better.
Here are a few more specific comments. Note that I have not commented on grammar here.
Intro Line 1: Except for Yelp, none of the location based social networks mentioned here are popular anymore. Instead, Twitter and Facebook are incorporating more location-based features. Foursquare is now a location-based service provider rather than a social network.
Intro: A decent overview of literature is provided.
Methods: The steps of clustering and CNN is well explained with illustrations. A panel including the entire algorithm would aid bring all elements together.
Dataset definition should go in Methods. Also, mention the time period of the dataset. What is ‘Social’ in the table?
Performance metrics should also go in Methods. Sensitivity and Specificity would also aid the paper.
Many parts of experimental data should also go in methods. Results should only include graphs and stats. The algorithms with which RecPOI is compared need to be discussed in Introduction.
A Discussion is needed. What does this achieve? What are the best use cases? On what data does it work better.
Author Response
RecPOID: POI Recommendation with Friendship aware and Deep CNN
Journal of Future Internet
Reviewer #2:
- Extensive editing of English language and style required.
RESPONSE: We are grateful to you for your insightful and helpful comments. We have revised the paper according to the comments. We tried to edit the article's English language grammar and style; we hope it will be accepted by you.
- Intro Line 1: Except for Yelp, none of the location based social networks mentioned here are popular anymore. Instead, Twitter and Facebook are incorporating more location-based features. Foursquare is now a location-based service provider rather than a social network.
RESPONSE: Thank you for mentioning these contents. According to your guidance, the first paragraph sentences of the introduction have been changed as follows:
Due to the explosive growth of smartphones, Location-based social networks (LBSNs) such as Yelp and Location-based service providers like Foursquare are popular. By 2020, Foursquare has over 55 million users per month and over 3 billion monthly visits to various locations worldwide; the Swarm app has 9 million check-ins per day.
- Methods: The steps of clustering and CNN is well explained with illustrations. A panel including the entire algorithm would aid bring all elements together.
RESPONSE: Due to your very good guidance, Algorithm 1(subsection 2.3) summarizes the detailed steps of the RecPOID algorithm.
- Dataset definition should go in Methods. Also, mention the time period of the dataset. What is ‘Social’ in the table?
RESPONSE: Based on your valued and effective points, the Datasets (subsection 2.4) has been moved, and the datasets time period is added as follows:
Yelp dataset challenge round 7 (access date: Feb 2016).
Gowalla dataset (access date: Feb. 2009 - Oct. 2010).
‘Social' refers to social relations, a user-user relationship, and is expected to share routine interest choices. It is a significant feature in POI recommendations.
- Performance metrics should also go in Methods. Sensitivity and Specificity would also aid the paper.
RESPONSE: Thank you for suggesting these tips, the Performance metrics (subsection 2.5) has been moved. Sensitivity and Specificity were evaluated. Their explanations have been added in the Performance Metrics (subsection 2.5) and the Discussion on Performance (subsection 3.2).
- Many parts of experimental data should also go in methods. Results should only include graphs and stats. The algorithms with which RecPOI is compared need to be discussed in Introduction.
RESPONSE: Based upon your helpful points, several subsections were moved into the Method section.
The Results section structure (subsection 3.1) has been changed, and the unit of Discussion on Performance (subsection 3.2) has been added.
According to your guidance, the algorithms compared with our approach are described in the introduction.
- A Discussion is needed. What does this achieve? What are the best use cases? On what data does it work better.
RESPONSE: Thank you for proposing these items, the Discussion on Performance (subsection 3.2) is added. The issues were discussed in various parts of the discussion, including:
In summary, this research’s contributions Propose a novel deep learning method to recommend an accurate sequence of top-k POIs to users considering a similar pattern’s friendship. The proposed RecPOID has improved the accuracy of POI recommendations compared to other state-of-the-art methods.
The two evaluated data sets have different data sparsity. The evaluation results show that our proposed algorithm works better on the Gowalla dataset than Yelp, and the friend importance in Gowalla is more influential than Yelp. It means there are richer social relationships in Gowalla and more under influence of their friends' check-ins.
Finally, we would like to thank you again for the excellent comments and suggestions
for improving the paper.
Author Response File: 
Author Response.pdf
Round 2
Reviewer 1 Report
Dear Authors,
I find the conducted major revision sufficient. Good job.
Please briefly address text formatting to be in accordance with the journal's guidelines (where and if applicable).
Kind regards,
Reviewer
Reviewer 2 Report
The authors have sufficiently addressed my concerns in this draft. The authors have done due diligence in improving all aspects of the paper and the new organization adds more clarity and information.
However, another round of editing for grammar would help.
