All reviews of published articles are made public. This includes manuscript files, peer review comments, author rebuttals and revised materials. Note: This was optional for articles submitted before 13 February 2023.
Peer reviewers are encouraged (but not required) to provide their names to the authors when submitting their peer review. If they agree to provide their name, then their personal profile page will reflect a public acknowledgment that they performed a review (even if the article is rejected). If the article is accepted, then reviewers who provided their name will be associated with the article itself.
Dear Authors,
Your article is Accepted. While in production please improve the minor English language corrections. Thank you.
[# PeerJ Staff Note - this decision was reviewed and approved by Claudio Ardagna, a PeerJ 'Computer Science' Section Editor covering this Section #]
The authors have addressed adequately my previous concerns on the experimental results. Once clarified, I consider this paper can be accepted for publication.
Additional and clear clarifications have been provided.
No further comments.
The work presents a procedure for enhancing neural collaborative filtering using hybrid feature selection for a recommendation. The study is interesting and well-written. The coherence and general flow are fine, apart from occasional difficulty due to the use of a few complex sentences and limited syntax usage.
The methodological design is easy to follow and the mathematics are quite easy to follow. However, the algorithmic flowcharts are missing, but these do not impact the flow of the work.
The findings were fairly reported and the results support the conclusions.
One of the reviewers commented on the quality of the graphics. It appears that the graphs still suffer from poor presentation. However, it was quite easy to understand the results.
Dear Authors,
Please revise and resubmit your manuscript. Thank you.
[# PeerJ Staff Note: Please ensure that all review comments are addressed in a rebuttal letter and any edits or clarifications mentioned in the letter are also inserted into the revised manuscript where appropriate. It is a common mistake to address reviewer questions in the rebuttal letter but not in the revised manuscript. If a reviewer raised a question then your readers will probably have the same question so you should ensure that the manuscript can stand alone without the rebuttal letter. Directions on how to prepare a rebuttal letter can be found at: https://peerj.com/benefits/academic-rebuttal-letters/ #]
The manuscript deals with an enhanced version of neural collaborative filtering for recommender systems which uses convolution instead of simple matrix factorization as a way to capture relationships between users and items. This brings an improved performance compared to previous implementations of neural CF.
The paper is clearly written, in a professional and academic language. The problem addressed in the paper is well motivated, and the bibliographic references are relevant and up to date too. The structure of the article is also clear, logical and easy to follow, and the graphics and figures are useful and informative. The datasets used in th study are public and accessible.
The proposed technique is clearly described, except for one point hat the authors should explain better: the embedding of users and items into features is not fully and completely specified. Since this is an important point for understanding the system design, the authors are encouraged to fix this part. The rest of the system is presented in a clear way.
The experimental result and the presentation of the results conforms to scholar standards and is well done: the datasets used for the experiments are significant and real, the performance metrics are clearly defined (and are typical in this field of study), and overall the numerical experiment have been created to find out the responses to the research questions posed in this paper. In this respect, the paper is absolutely correct.
This is, on the other hand, replicable research (provided the explanation of the embeddings pointed out previously is solved).
There is, however, some margin for improvements in the presentation:
1) The figures and performance plots do not have a high quality. It is suggested to improve them, both in quality and in clarity.
2) According to Figure 1, there seems to be no interaction in the computation between the user features and the item features. Please, explain this discrepancy.
3) The results with the proposed technique show also perfect prediction (roughly 99%) for HR, which is much better than with other methods. This sharp improvement should be explained and discussed better in the paper. Could there be some form of overfitting in the proposed system?
While the technique can be reproduced and tested again, such remarkable performance improvement must be explained directly by the authors, particularly the reasons for achieving those excellent results.
The research paper is well presented and well structured. The authors realized a new contribution in the domain of the collaborative filtering recommendation.
The authors should add some new references, the most updated one is 2020.
The results of the proposed model should be improved by its comparison with other models in terms of Within-Cluster Sum of Square, Silhouette coefficient, V-measure and Fowlkes-Mallows Scores
The proposed model is well presented and is useful to increase the understanding and the interpretation of the input features through the interaction maps and the correlation between items. Furthermore, the use of the Generalized Matrix Factorization a solution for preventing the overfitting issue.
The evaluation of the proposed model focused on the recommendation ranking using the hit ratio and the normalized discounted cumulative gain which are evaluation protocols in the recommendation domain. It will be more convenient if the authors show the efficiency of the model by measuring the Within-Cluster Sum of Square, Silhouette coefficient, V-measure and Fowlkes-Mallows Scores
Although the authors presented a link that contains the datasets, it will be better to present a paragraph to describe the features of each dataset.
All text and materials provided via this peer-review history page are made available under a Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.