Enhancing Recommendation Systems through SVD-based collaborative filtering and community detection.

Skip to main content An official website of the United States government Here's how you know Here's how you know Official websites use .gov A .gov website belongs to an official government organization in the United States. Secure .gov websites use HTTPS A lock ( Lock Locked padlock icon ) or https:// means you've safely connected to the .gov website. Share sensitive information only on official, secure websites. Search Log in Dashboard Publications Account settings Log out Search… Search NCBI Primary site navigation Search Logged in as: Dashboard Publications Account settings Log in Search PMC Full-Text Archive Search in PMC Journal List User Guide PERMALINK Copy As a library, NLM provides access to scientific literature. Inclusion in an NLM database does not imply endorsement of, or agreement with, the contents by NLM or the National Institutes of Health. Learn more: PMC Disclaimer | PMC Copyright Notice PLoS One . 2026 Apr 13;21(4):e0346579. doi: 10.1371/journal.pone.0346579 Search in PMC Search in PubMed View in NLM Catalog Add to search Enhancing Recommendation Systems through SVD-based collaborative filtering and community detection T Keerthika T Keerthika 1 Amrita School of Artificial Intelligence, Amrita Vishwa Vidyapeetham, Coimbatore, Tamil Nadu, India Writing – original draft Find articles by T Keerthika 1 , Rajathi G Ignisha Rajathi G Ignisha 2 Manipal Institute of Technology Bengaluru, Manipal Academy of Higher Education, Manipal, India Conceptualization, Methodology Find articles by Rajathi G Ignisha 2, * , Vedhapriyavadhana Rajamani Vedhapriyavadhana Rajamani 3 School of Computing, Engineering and Physical Sciences, University of the West of Scotland, London, England, United Kingdom Conceptualization, Writing – review & editing Find articles by Vedhapriyavadhana Rajamani 3 , Surya Santhosh Kumar Surya Santhosh Kumar 1 Amrita School of Artificial Intelligence, Amrita Vishwa Vidyapeetham, Coimbatore, Tamil Nadu, India Investigation, Software Find articles by Surya Santhosh Kumar 1 , Karthigai Selvam Karthigai Selvam 1 Amrita School of Artificial Intelligence, Amrita Vishwa Vidyapeetham, Coimbatore, Tamil Nadu, India Resources, Validation Find articles by Karthigai Selvam 1 , M R Aiyyappan M R Aiyyappan 1 Amrita School of Artificial Intelligence, Amrita Vishwa Vidyapeetham, Coimbatore, Tamil Nadu, India Data curation, Investigation Find articles by M R Aiyyappan 1 , L S Thoshi Babu L S Thoshi Babu 1 Amrita School of Artificial Intelligence, Amrita Vishwa Vidyapeetham, Coimbatore, Tamil Nadu, India Visualization, Writing – original draft Find articles by L S Thoshi Babu 1 Editor: Shih-Lin Lin 4 Author information Article notes Copyright and License information 1 Amrita School of Artificial Intelligence, Amrita Vishwa Vidyapeetham, Coimbatore, Tamil Nadu, India 2 Manipal Institute of Technology Bengaluru, Manipal Academy of Higher Education, Manipal, India 3 School of Computing, Engineering and Physical Sciences, University of the West of Scotland, London, England, United Kingdom 4 National Changhua University of Education, TAIWAN Competing Interests: The authors have declared that no competing interests exist. ✉ * E-mail: [email protected] Roles T Keerthika : Writing – original draft Rajathi G Ignisha : Conceptualization, Methodology Vedhapriyavadhana Rajamani : Conceptualization, Writing – review & editing Surya Santhosh Kumar : Investigation, Software Karthigai Selvam : Resources, Validation M R Aiyyappan : Data curation, Investigation L S Thoshi Babu : Visualization, Writing – original draft Shih-Lin Lin : Editor Received 2025 Nov 26; Accepted 2026 Mar 21; Collection date 2026. © 2026 Keerthika et al This is an open access article distributed under the terms of the Creative Commons Attribution License , which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. PMC Copyright notice PMCID: PMC13075704  PMID: 41973699 Abstract The recommendation systems often face challenges like low data density, scalability issues and absence of interpretability, whereas classical Collaborative Filtering (CF) which is based on Singular Value Decomposition (SVD) shows support by being scalable, weakened frequently in circumstances of extreme sparsity. Conversely, Graph Neural Networks (GNNs) are very accurate yet do not tend to have explanatory power. In a novel way, this research presents a hybrid framework that is a sequential combination of Louvain community identification using SVD-based collaborative filtering to overcome the sparsity-interpretable trade-off. It is unlike the existing models that utilize communities only, to pre-partition the user space, modularity-based clustering is employed to regularize it, enabling SVD to act on more dense homogeneous sub-matrices. This methodological contribution is a very useful way to cut down on computational noise and overhead and to make the community-level justifications of recommendations. In the experimental analysis, the Netflix Prize data set produced a Root-Mean-Square Error (RMSE) of 0.9966, a Mean value of 0.9966 and an Absolute Error (MAE) of 0.7968. This hybrid model achieves competitive predictive performance with significantly higher interpretability and lower computational cost than complex deep learning baselines, despite a modestly higher RMSE due to the deliberate trade-off for transparency and efficiency on extremely sparse data. The framework enables scalable and transparent recommendation engines suitable for large-scale sparse datasets. Introduction The present-day Recommendation Systems (RS) are based on the old recommendations techniques that included Content-Based Filtering (CBF) and Collaborative Filtering (CF) [ 1 , 2 ]. The desire of the users converges in similar preferences when they have overlap in their past behaviors based on the CF principles. The CF system works under two different subcategories that either form a recommendation on similar behavior of the user or form a recommendation by similarity of items to form suggestions [ 2 , 3 ]. The weaknesses of CF to process the hidden item-user relations are usually constrained by the failure to handle data sparsity and scalability problems [ 4 ]. CBF has limited capabilities to discover implicit patterns and extrapolate knowledge beyond specified features due to the nature of its recommendation capabilities being dependent on the characteristics or profile of items to interpret recommendations but has been shown to be incompetent to do so [ 2 ]. Singular Value Decomposition (SVD) has been recognized as an important component of Matrix Factorization (MF) methods that have achieved some potent results in the past with the assistance of Singular Value Decomposition (SVD) [ 5 ]. SVD is used to obtain latent factors by decomposing the user-item interaction matrix, which expose underlying preferences as well as characteristics. This approach reduces the complexity of data and also identifies relationships that none of the other CF solutions would have identified before [ 6 ]. The use of SVD in improving the recommendation accuracy is however limited. The first is when the datasets used are very sparse, when the system does not identify internal links between users and form communities, and in cold-start situations, which is disfavored in that case, as well as with very sparse datasets, in this instance, the system is unable to recognize internal relationships between users that form communities [ 4 , 7 , 8 ]. In the recent developments, there has been a transition to deep learning and Graph Neural Networks (GNNs) [ 9 , 10 ]. Most recently, Transformer-based models and Large Language Models (LLMs) have been studied in the context of sequential recommendation, with state-of-the-art accuracy, but typically high computational latency and black-box proprietary qualities. Despite these advances, the industry still requires the solutions that will strike a balance between high accuracy and interpretability and low resources consumption. This paper to fill these gaps revisits the prism of community-detection of matrix factorization. Latent patterns are identified with the help of Singular Value Decomposition (SVD) and natural group structures are employed with the help of Louvain community detection. [ 7 , 11 ]. Contributions The main contributions of present research are the following: Hyperscale Hybrid Architecture: It offers a progressive approach to compute Louvain community detection with SVD, making the change to local factorization of matrices. This solves the problem of data sparsity by forming more dense user clusters. Strength of Interpretability: As opposed to black-box GNN models, the proposed approach can give transparent, communal explanations behind recommendations (e.g., peer-group influence). Scalable Performance: It is shown that the user space partitioning limits the dimensionality load on SVD and competitive accuracy (RMSE 0.9966) is achieved with reduced computational overhead than deep learning baselines. Organization of the paper The remainder of this paper is organized as follows: Section 2 reviews related literature on the topics of collaborative filtering and community detection. Section 3 elaborates the methodology framework, mathematical equations, theoretical assumptions and complexity analysis. Section 4 describes the proposed hybrid architecture, hyperparameter finding and experimental configuration. Section 5 contains the findings of the experiment and a comparative discussion. Finally, section 6 concerns the conclusion of the research and gives the direction of the future research. Related work Recommendation systems development has weakened as a result of the integration of graph-based approaches, community detection applications, as well as the SVD method of matrix factorization. Detailed analysis has shown that the graph-based recommender systems are meaningful since they generate key information using graph representations to form powerful recommendations and enhance readability [ 1 ]. Graph-based recommenders have been regarded as transparent to research the effects of graph structures on trust in a recommendation, such as detecting communities and modeling nodes [ 12 ].Some of the publications suggest that the methodologies of graph learning can be used to address the issue of data sparsity and cold-start successfully. [ 4 , 13 ]. The division process that the community detection algorithms perform on the sets of users forms unique groups, which results in better individualized recommendation outputs. There are three most popular community detection algorithms: Louvain, Leiden, and Label Propagation helping extensive network applications [ 2 , 11 ]. The Louvain method has been one of the favorite strategies to create the best-quality community structures, and hence it is adopted in several practical cases of the real world contexts as well [ 9 , 14 ]. The last memory efficient advances have boosted the scalability of such algorithms in large scale database operations [ 15 ]. Recommender systems that appeal combine collaborative and content based methods and are being popular because they offer high accuracy and diversity on the system. Study of SVD in collaborative filtering has grown because the algorithm identifies concealed patterns in user-item data matrices. The study of SVD in collaborative filtering has been broadened since the algorithm detects hidden patterns in matrices of user-item data sets. Hybrid recommendation systems derive the advantages of two methods to address sparsity constraints and overfitting and enhance user-specific recommendations. SVD based systems predict user preference with a robust performance by incorporating bias terms, which execute user, and item based methods of behaving in their selection process of a product or service [ 6 , 7 ]. Graph neural networks (GNNs) form a core field of graph representations that combine the information of each node in a manner that allows training suggestions. [ 9 , 16 , 17 ]. The studies mainly explore the methods used to simulate the interaction of the users with various items. As an example, Multi-Behavior GNN (MB-GNN) is an improvement to representation learning by processing the various types of interaction, including clicks and purchases [ 10 ]. Concluding remarks on research gap The gaps in literature, through which the effectiveness of the balance between the interpretability and sparsity management could be achieved, are identified. As SVD can help to scale and GNNs can help to improve accuracy, not many combined models can use the modularity of community detection to explicitly clean up the interaction matrix first before factorization. The proposed work aims at bridging this gap by proposing a sequential framework that would use community detection, not only to regularize the data, but also to pre-process the data in order to generate dense, homogenous sub-matrices where simpler factorization is possible. Methodology Singular Value Decomposition (SVD) Singular Value Decomposition (SVD) is a classical element of collaborative filtering, and it provides an excellent mathematical backbone to identify latent variables used to predict user-item interactions. The user-item interaction data is represented by a matrix, R . SVD solves the problem of sparsity by breaking down the matrix R into three components of the form of U , Σ and V T as below mathematical expression: R = U Σ V ⊤ , (1) Here, U is an m × k matrix capturing latent features of users, Σ is a k × k diagonal matrix including singular values, and V T is a k × n matrix representing latent features of items. The optimization process finds minimal squared differences between user ratings predictions and actual values. The optimization includes the following formula: ℒ = ∑ ( u , i ) ∈ Ω ( r u , i − r ^ u , i ) 2 + λ ( ‖ p u ‖ 2 + ‖ q i ‖ 2 ) (2) where Ω symbolizes the collection of user-item interactions, r u , i is the actual rating, and r ^ u , i is the predicted rating computed as: r ^ u , i = μ + b u + b i + p u ⊤ q i , (3) Louvain community-detection method The core function of the Louvain algorithm involves user preference cluster discovery. The strength of network division is quantified through Modularity Q , defined as [ 14 , 15 ]: Q = 1 2 m ∑ i , j [ w i j − k i k j 2 m ] δ ( c i , c j ) , (4) where w ij represents the weight of the edge between nodes i and j , k i and k j represent degrees, and δ ( c i , c j ) is 1 if nodes are in the same community. Assumptions and theoretical constraints Singular Value Decomposition (SVD) Singular Value Decomposition (SVD) is a classical element of collaborative filtering, and it provides an excellent mathematical backbone to identify latent variables used to predict user-item interactions. The data of interaction between the user and the item are represented in a form of a matrix, R . SVD resolves the sparsity issue by decomposing the matrix R into three terms, namely X , Σ and Y ⊤ , as follows in the mathematical expression: R = X Σ Y ⊤ , (5) Here, X is an m × p matrix capturing latent features of users, Σ is a p × p diagonal matrix including singular values, and Y T is a p × n matrix representing latent features of items. The optimization process identifies small squared differences between the prediction of the user ratings and the actual ones. The optimization contains the following formula: ℒ = ∑ ( x , i ) ∈ Ω ( r x , i − r ^ x , i ) 2 + λ ( ‖ p x ‖ 2 + ‖ q i ‖ 2 ) (6) with the representation of the set of interactions between users and items in the form of the shortcut Ω , and the real rating on the interaction between a user and item denoted by r x , i , the predicted rating denoted by r ^ x , i , and calculated as follows: r ^ x , i = μ + b x + b i + p x ⊤ q i , (7) Louvain community-detection method The core function of the Louvain algorithm involves user preference cluster discovery. The strength of network division is quantified through Modularity Q , defined as [ 14 , 15 ]: Q = 1 2 m ∑ i , j [ w i j − p i p j 2 m ] δ ( c i , c j ) , (8) where p i and p j denote degrees, w ij is the weight of the edge between nodes i and j , and δ ( c i , c j ) is 1 if nodes are in the same community. Assumptions and theoretical constraints The proposed hybrid system will be performed based on certain theoretical assumptions that SVD and Louvain algorithms should work with appropriately: Low-Rank Assumption: It is assumed that the user-item interaction matrix is low-rank ( p ≪ min ( m , n ) ) user preferences can be well modeled by a few latent variables. Homophily Assumption Homophily assumption: Louvain method is premised on the fact that the users within the same community are statistically important with regard to their taste preferences compared to the who were not a part of the community. Identifiability with SVD: To achieve unique reconstruction of the rating matrix (up to permutation), one assumes that the singular values are distinct and that intra-community sub-matrices, while not fully dense, are sufficiently connected to ensure convergence during Stochastic Gradient Descent (SGD). Computational complexity analysis One of the main strengths of the suggested strategy is scalability. It is a complexity that comprises two parts: Louvain Algorithm :Community detector has time bound of O ( N log N ) , where N is the number of users. It is an amazing invention with sparse graphs. Local SVD: Standard global SVD has a complexity of O ( p · | R | · I ) , where p is latent dimensions, | R | is the number of ratings, and I is iterations. By partitioning users into C communities, C independent SVDs are performed on smaller matrices. While the total operations remain proportional, the convergence time is reduced because the sub-matrices are denser and more homogenous, requiring fewer iterations I to minimize error. The proposed method is much lighter in terms of memory and floating-point operations as compared to GNN-based methods (e.g., NGCF) which scale at O ( L · N · d 2 + L · | R | ) . Proposed method The proposed hybrid recommendation system brings together Singular Value Decomposition algorithm with Louvain community detection to embrace group-level dependencies while addressing sparsity issues as shown in Fig 1 . The first stage establishes an R user-item interaction matrix. The next process will be to construct a user similarity graph that represents the relationship between users considering their interaction profile. The following formula is computed to compute the similarity of the users in terms of cosine similarity when comparing two users, x and y : Fig 1. Architectural workflow of the proposed Hybrid SVD-Louvain Recommendation System. Open in a new tab sim ( x , y ) = ∑ i ∈ I R x , i R y , i ∑ i ∈ I R x , i 2 ∑ i ∈ I R y , i 2 (9) Once communities are identified, SVD is applied within each community. For each community, a submatrix R c is extracted. SVD is then performed on R c to decompose it: R c ≈ P c Σ c Q c T (10) Hyperparameter tuning To ensure optimal performance ( Equations 4 and 5 ), a Grid Search strategy was employed for hyperparameter optimization. Learning Rate ( α ): Values were explored in the range {0.001, 0.005, 0.01, 0.02}. It was observed that α > 0.02 led to divergent loss, while α < 0.001 resulted in slow convergence. The optimal value was identified as 0.005. Regularization ( λ ): Tests were conducted for λ ∈ { 0.01 , 0.02 , 0.05 , 0.1 } . Values lower than 0.02 caused overfitting on the training communities, while λ > 0.1 reduced predictive accuracy by penalizing latent features too heavily. The value λ = 0.02 was selected for the reported results. Results and discussion A thorough evaluation occurred through analysis of the Netflix Prize dataset. The dataset contains more than 24 million ratings from about 470,758 users ( Table 1 ). Table 1. Dataset statistics from the Netflix Prize (Kaggle). Dataset Users Movies Total Ratings Rating Range Sparsity Netflix Prize (Kaggle) 470758 4499 24053764 1–5 0.9886 Open in a new tab Fig 2 shows that the distribution of the ratings has a long-tail, which proves the sparseness of the data that prompts the need to adopt the community-based approach. Fig 2. Distribution of ratings for movies across the dataset. Open in a new tab A random selection of 100 users and 50 movies was used to do a detailed analysis. Community detection algorithms on the Louvain community on the data set identified six distinct clusters within the data set. Fig 3 illustrates the network topology, with Community 0 being shown as densely connected than Community 2 which is sparsely connected. This graphic distinction validates the effectiveness of the algorithm to divide separate preferences groups. Fig 3. Network visualization of six detected user communities based on movie rating patterns. Open in a new tab Fig 4 presents the distribution of users with Community 0 having the largest number of users (25) which can be interpreted as a prevailing pattern in the sample in terms of preference whereas other communities are niche interests. Fig 4. Distribution of users in each of the detected communities. Open in a new tab Fig 5 illustrates a heatmap of the community-movie matrix. Indicatively, films that prove to be highly ranked in Community 1 possess clearly lesser affinity in Community 3, confirming the preference segregation that was obtained by the model ( Table 2 ). Fig 5. Heatmap of the community-movie matrix for the 15 highest-rated movies. Open in a new tab Table 2. Results for Proposed Hybrid Model on test set. Metric Fold 1 Fold 2 Fold 3 Mean Std. Deviation RMSE (test set) 0.9957 0.9943 0.9998 0.9966 0.0023 MAE (test set) 0.7970 0.7959 0.7975 0.7968 0.0007 Open in a new tab Comparative analysis and discussion To assess the significance of the obtained RMSE (0.9966), a comparison was made against standard values reported in literature. The hybrid SVD-Louvain approach outperforms standard memory-based CF and performs competitively with basic matrix factorization baselines. Fig 6 visualizes the performance gap. Although GNN-based methods achieve slightly lower RMSE (approx 0.92) [ 17 ], this difference arises because the proposed method prioritizes interpretability (clear community-based explanations) and computational efficiency (local factorization on denser sub-matrices) over marginal accuracy gains from black-box GNN architectures, which require substantially higher resources on sparse datasets like Netflix. Consequently, the proposed method (RMSE 0.9966) significantly outperforms isolated SVD (typically 1.05) [ 6 ] while retaining strong interpretability. This trade-off is quantified in Table 3 . Fig 6. RMSE Performance Comparison: Isolated SVD vs. GNN vs. Proposed Method. Open in a new tab Table 3. Comparison of Proposed Hybrid Method vs. SVD and GNN Benchmarks. Methodology Typical RMSE Sparsity Handling Interpretability Comp. Cost Isolated SVD 1.00–1.05 [ 6 ] Low Moderate Low GNN-based (e.g., NGCF) 0.90–0.95 [ 17 ] High Very Low (Black Box) High Proposed Hybrid 0.9966 High High (Community) Moderate Open in a new tab Discussion on diversity and personalization Although the key measures are error-based (RMSE/MAE), implicitly, the hybrid architecture of the architecture increases the diversity of the recommendations. The localization of collaborative filtering to local communities eliminates the so-called popularity bias (popular globally crowd out unpopular items in the globally popular item list). In smaller, niche communities, like “Indie Horror Fans” users get recommendations tailored to the interests of their group, instead of generic blockbusters, which in itself enhances the catalog coverage and diversity of personalization than a global SVD model. Conclusion In this paper, a hybrid recommendation system was created and justified by combining the Louvain community detection algorithm with Singular Value Decomposition (SVD). The data sparseness inherent in global matrix factorization was overcome by splitting the user-item interaction space into different communities, which was determined by modularity. This community-centered factorization performed in experimentation showed a competitive RMSE of 0.9966. Most importantly, it was determined that the use of SVD in the framework of localized groups of users preserved predictive power but greatly improved interpretability in comparison to the black-box deep learning options. Although these are the benefits, the suggested approach has its shortcomings. The use of pre-computed communities implies that real-time updates of incoming users (cold-start) are based on periodic re-clustering of a group and could be computationally expensive. Future studies ought to consider injecting in temporal dynamics, to simulate the dynamics of the user community memberships with time. Moreover, information related to user reviews or social network links might be introduced, and this addition would likely help to perfect the community detection process. It is further suggested that this hybrid architecture should also be tested on a distributed computing framework to determine the feasibility of the recommendation in real-time on a large scale when dealing with the ultra-large e-commerce platform. Data Availability All relevant data are within the paper. Further the dataset link is provided in the manuscript. The dataset can be accessed from the Kaggle repository at: { https://www.kaggle.com/datasets/netflix-inc/netflix-prize-data }. The dataset from the original is reproduced in the following dataset doi: 10.6084/m9.figshare.31428857 . Funding Statement The author(s) received no specific funding for this work. References 1. Zhang J, Fei J, Song X, Feng J. An Improved Louvain Algorithm for Community Detection. In: 2021. 10.48550/arXiv.2110.00891 [ DOI ] 2. Zhao Y, Zhou C, Cao J. Building a Movie Recommendation System Using SVD Algorithm. Scientific Reports. 2024;14:3853. doi: 10.1038/s41598-024-54612-338360899 [ DOI ] [ Google Scholar ] 3. Chen L, Zhang W. Movie recommendation system based on collaborative filtering and matrix factorization. Journal of Emerging Technologies and Innovative Research. 2023. 4. Choudhary C, Singh I, Kumar M. Community detection algorithms for recommendation systems: techniques and metrics. Computing. 2022;105(2):417–53. doi: 10.1007/s00607-022-01131-z [ DOI ] [ Google Scholar ] 5. Soman ST, Soumya SVJ, Soman KP. Singular Value Decomposition: A Classroom Approach. International Journal of Recent Trends in Engineering. 2009;2(1). [ Google Scholar ] 6. Koren Y, Bell R, Volinsky C. Matrix Factorization Techniques for Recommender Systems. Computer. 2009;42(8):30–7. doi: 10.1109/mc.2009.263 [ DOI ] [ Google Scholar ] 7. Blondel VD, Guillaume J-L, Lambiotte R, Lefebvre E. Fast unfolding of communities in large networks. J Stat Mech. 2008;2008(10):P10008. doi: 10.1088/1742-5468/2008/10/p10008 [ DOI ] [ Google Scholar ] 8. Jothilakshmi SL, Bharathi R. Survey on Collaborative Filtering Technique for Recommender System Using Deep Learning. Lecture Notes in Electrical Engineering. Springer Nature Singapore. 2023. p. 217–25. 10.1007/978-981-19-7169-3_20 [ DOI ] [ Google Scholar ] 9. Gao C, Zheng Y, Li N. A survey of graph neural networks for recommender systems: challenges, methods, and directions. arXiv preprint. 2022. 10.48550/arXiv.2109.12843 [ DOI ] 10. Zhu H, Kapoor V, Sharma P. Reviewing developments of graph convolutional network techniques for recommendation systems. 2023. 10.48550/arXiv.2311.06323 [ DOI ] 11. Gasparetti F, Sansonetti G, Micarelli A. Community Detection in Social Recommender Systems: A Survey. Applied Intelligence. 2021;51(6):3975–95. doi: 10.1007/s10489-020-02162-2 [ DOI ] [ Google Scholar ] 12. Li M, Zhao L, Ren Y. Graph Neural Network Recommendation Algorithm Based on Improved Dual-Tower Model. Gels. 2024;11(2):141. doi: 10.3390/gels11020141 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 13. Wu S, Sun F, Zhang W. Graph Neural Networks in Recommender Systems: A Survey. ACM Computing Surveys. 2022;55(7):1–37. doi: 10.1145/3568022 [ DOI ] [ Google Scholar ] 14. Hamilton WL, Ying R, Leskovec J. Inductive Representation Learning on Large Graphs. In: Advances in Neural Information Processing Systems (NeurIPS), 2017. https://doi.org/arXiv:1706.02216 15. Mohammadi M, Fazlali M, Hosseinzadeh M. Parallel Louvain Community Detection Algorithm Based on Dynamic Thread Assignment on Graphic Processing Unit. Journal of Electrical and Computer Engineering Innovations (JECEI). 2022;10(1):75–88. doi: 10.22068/JEECI.10.1.75 [ DOI ] [ Google Scholar ] 16. Yang L, Wang Y, Tang J. Graph Neural Networks in Recommender Systems: A Survey. 2020. 10.48550/arXiv.2011.02260 [ DOI ] 17. Sharma K, Lee YC, Nambi S, Park Y. A survey of graph neural networks for social recommender systems. ACM Computing Surveys. 2024. 10.1145/3637528 [ DOI ] PLoS One. doi: 10.1371/journal.pone.0346579.r001 Decision Letter 0 Shih-Lin Lin Shih-Lin Lin Academic Editor Find articles by Shih-Lin Lin Author information Copyright and License information Roles Shih-Lin Lin : Academic Editor © 2026 Shih-Lin Lin This is an open access article distributed under the terms of the Creative Commons Attribution License , which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. PMC Copyright notice 9 Dec 2025 Dear Dr. Rajathi G, Thank you for submitting your manuscript to PLOS ONE. After careful consideration, we feel that it has merit but does not fully meet PLOS ONE’s publication criteria as it currently stands. Therefore, we invite you to submit a revised version of the manuscript that addresses the points raised during the review process. Please submit your revised manuscript by Jan 23 2026 11:59PM. If you will need more time than this to complete your revisions, please reply to this message or contact the journal office at [email protected] . When you're ready to submit your revision, log on to https://www.editorialmanager.com/pone/ and select the 'Submissions Needing Revision' folder to locate your manuscript file.. When you're ready to submit your revision, log on to https://www.editorialmanager.com/pone/ and select the 'Submissions Needing Revision' folder to locate your manuscript file.. When you're ready to submit your revision, log on to https://www.editorialmanager.com/pone/ and select the 'Submissions Needing Revision' folder to locate your manuscript file.. When you're ready to submit your revision, log on to https://www.editorialmanager.com/pone/ and select the 'Submissions Needing Revision' folder to locate your manuscript file. A rebuttal letter that responds to each point raised by the academic editor and reviewer(s). You should upload this letter as a separate file labeled 'Response to Reviewers'. A marked-up copy of your manuscript that highlights changes made to the original version. You should upload this as a separate file labeled 'Revised Manuscript with Track Changes'. An unmarked version of your revised paper without tracked changes. You should upload this as a separate file labeled 'Manuscript'. If you would like to make changes to your financial disclosure, please include your updated statement in your cover letter. Guidelines for resubmitting your figure files are available below the reviewer comments at the end of this letter. If applicable, we recommend that you deposit your laboratory protocols in protocols.io to enhance the reproducibility of your results. Protocols.io assigns your protocol its own identifier (DOI) so that it can be cited independently in the future. For instructions see: https://journals.plos.org/plosone/s/submission-guidelines#loc-laboratory-protocols . Additionally, PLOS ONE offers an option for publishing peer-reviewed Lab Protocol articles, which describe protocols hosted on protocols.io. Read more information on sharing protocols at . Additionally, PLOS ONE offers an option for publishing peer-reviewed Lab Protocol articles, which describe protocols hosted on protocols.io. Read more information on sharing protocols at . Additionally, PLOS ONE offers an option for publishing peer-reviewed Lab Protocol articles, which describe protocols hosted on protocols.io. Read more information on sharing protocols at . Additionally, PLOS ONE offers an option for publishing peer-reviewed Lab Protocol articles, which describe protocols hosted on protocols.io. Read more information on sharing protocols at https://plos.org/protocols?utm_medium=editorial-email&utm_source=authorletters&utm_campaign=protocols .... We look forward to receiving your revised manuscript. Kind regards, Shih-Lin Lin, Ph.D Academic Editor PLOS One Journal Requirements: When submitting your revision, we need you to address these additional requirements. 1. Please ensure that your manuscript meets PLOS ONE's style requirements, including those for file naming. The PLOS ONE style templates can be found at https://journals.plos.org/plosone/s/file?id=wjVg/PLOSOne_formatting_sample_main_body.pdf and https://journals.plos.org/plosone/s/file?id=ba62/PLOSOne_formatting_sample_title_authors_affiliations.pdf 2. Please note that PLOS One has specific guidelines on code sharing for submissions in which author-generated code underpins the findings in the manuscript. In these cases, we expect all author-generated code to be made available without restrictions upon publication of the work. Please review our guidelines at https://journals.plos.org/plosone/s/materials-and-software-sharing#loc-sharing-code and ensure that your code is shared in a way that follows best practice and facilitates reproducibility and reuse. 3. Please update your submission to use the PLOS LaTeX template. The template and more information on our requirements for LaTeX submissions can be found at http://journals.plos.org/plosone/s/latex .... 4. We note that your Data Availability Statement is currently as follows: All relevant data are within the manuscript and its Supporting Information files Please confirm at this time whether or not your submission contains all raw data required to replicate the results of your study. Authors must share the “minimal data set” for their submission. PLOS defines the minimal data set to consist of the data required to replicate all study findings reported in the article, as well as related metadata and methods ( https://journals.plos.org/plosone/s/data-availability#loc-minimal-data-set-definition ). For example, authors should submit the following data: - The values behind the means, standard deviations and other measures reported; - The values used to build graphs; - The points extracted from images for analysis. Authors do not need to submit their entire data set if only a portion of the data was used in the reported study. If your submission does not contain these data, please either upload them as Supporting Information files or deposit them to a stable, public repository and provide us with the relevant URLs, DOIs, or accession numbers. For a list of recommended repositories, please see https://journals.plos.org/plosone/s/recommended-repositories . If there are ethical or legal restrictions on sharing a de-identified data set, please explain them in detail (e.g., data contain potentially sensitive information, data are owned by a third-party organization, etc.) and who has imposed them (e.g., an ethics committee). Please also provide contact information for a data access committee, ethics committee, or other institutional body to which data requests may be sent. If data are owned by a third party, please indicate how others may request data access. 5. We note you have included a table to which you do not refer in the text of your manuscript. Please ensure that you refer to Table 1 in your text; if accepted, production will need this reference to link the reader to the Table. 6. If the reviewer comments include a recommendation to cite specific previously published works, please review and evaluate these publications to determine whether they are relevant and should be cited. There is no requirement to cite these works unless the editor has indicated otherwise. [Note: HTML markup is below. Please do not edit.] Reviewers' comments: Reviewer's Responses to Questions Comments to the Author 1. Is the manuscript technically sound, and do the data support the conclusions? Reviewer #1: Yes Reviewer #2: Partly Reviewer #3: Yes Reviewer #4: Yes ********** 2. Has the statistical analysis been performed appropriately and rigorously? -->?> Reviewer #1: Yes Reviewer #2: Yes Reviewer #3: Yes Reviewer #4: Yes ********** 3. Have the authors made all data underlying the findings in their manuscript fully available??> The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.--> Reviewer #1: Yes Reviewer #2: No Reviewer #3: Yes Reviewer #4: Yes ********** 4. Is the manuscript presented in an intelligible fashion and written in standard English??> Reviewer #1: Yes Reviewer #2: Yes Reviewer #3: Yes Reviewer #4: Yes ********** Reviewer #1: 1. The article aligns well with the journal's scope and is presented in an organized manner, making it easy to read and follow. 2. The authors have developed a hybrid system that combines SVD-based collaborative filtering and the Louvain community detection algorithm. 3. The proposed method applies SVD for dimensionality reduction and community detection methods to group users with similar preferences. 4. The analysis of the Netflix Prize dataset demonstrated the effectiveness of the proposed approach, yielding a root-mean-square error of 0.9966 and a mean absolute error (MAE) of 0.7968. 5. The authors concluded that these metrics demonstrate high performance compared to isolated SVD implementations and offer significantly better interpretability than GNN-based approaches. 6. The methodology is clearly presented, with extensive explanations and discussions of both the method and the results. However, I suggest the authors address the following issues: 1. While informative, the abstract could be more explicit about the research's key findings and contributions. Including specific results from the case study could enhance its impact. 2. The authors need to compare the performance of the proposed method with existing literature to demonstrate its robustness. 3. Subheadings directly after headings, as indicated in the attached file, are not preferred. Please add one or two sentences to introduce the section and keep the paper flowing. 4. I strongly encourage the authors to review the reference list for any duplicates carefully. I have noticed that References [12] and [28] are identical. Additionally, References [13], [18], and [27] are quite similar. Addressing these issues will improve the clarity and reliability of the work. However, references should be ordered by their first appearance in the manuscript. 5. I suggest the authors slightly revise the language of the article. The attached file includes many suggestions. Reviewer #2: 1. The abstract makes strong claims about improved interpretability and performance, but it does not clearly explain the novelty of the hybrid SVD–Louvain approach or how it differs from existing community-enhanced recommendation models. Adding a clearer statement of the methodological contribution would strengthen the abstract. 2. The description of experimental results lacks context—no baseline RMSE/MAE values, dataset details, or comparison conditions are provided—making it difficult for readers to judge the significance of the reported improvements. Including comparative benchmarks or percentage gains would increase clarity and impact.. 3. The introduction should clearly conclude with a distinct section highlighting the novel contributions of your work. 4. At the ending of the intro, it is advised to add a para that mentions briefly what each next section contains. 5. The literature review should benefit from more explorations of previous studies. 6. The discussion section needs to be expanded to more thoroughly analyze the results. 7. The first paragraph of the conclusion should succinctly summarize the contributions of the study in past tense. 8. The second paragraph of the conclusion should provide clear and actionable future recommendations. 9. Equations are not properly cited, please add original references. 10. Manuscript is highly low on visualizations. Reviewer #3: The paper introduces a hybrid recommendation system that combines a user item interaction matrix factorized by Singular Value Decomposition (SVD) with Louvain community detection on a user similarity graph constructed via cosine similarity. I have the following major comments: 1. The introduction can include more recent state of the art works. 2. Please specify the novelty and contributions of the work. 3. The assumptions under which the hybrid recommendation system properly functions along with possible identifiability SVD conditions should be mentioned. 4. In equations (4) and (5) it is not clear how to choose the learning rate and the regularization parameter. 5. It is not clear how diversity is treated, where the simulations is only in terms of RMSE, MAE, etc. 6. Simulations should show the impact of the learning rate and the regularization parameter on the obtained RMSE. 7. The authors should conduct a rigorous computational complexity analysis which was not performed. 8. The work should compare with more state of the art methods. 9. Please fix the text size of the figure, for example Fig. 1 has very small text. Please revise. Reviewer #4: 1) It is suggested to avoid the word “we”, “our” in the abstract and in other places of the manuscript. Write in passive voice. 2) The citations in the Introduction have referred randomly. It becomes problematic for the readers to follow the reference list. For example, the introduction is started with citations [5, 6]. It is not a correct way of citing literatures for a good technical article. It is suggested to rearrange the reference list as per their mentioning. 3) It is suggested to write concluding remarks at the end of the related work about what new thing has been focused in the proposed work which was being revealed from the literature survey as research gap. 4) Include a flowchart at the beginning of the section 4 to describe the step-by-step process of the proposed design. It will be very much helpful to the readers to understand the whole work at a glance. 5) Figure 1 is not mentioned in the text. It is to be mentioned. 6) Elaborate Figure 2, 3 and 4 in more details by choosing specific examples in the plots to validate this claim. 7) Results and analysis section needs improvement by including more simulation plots. 8) Include a comparative table at the end of the results and analysis section with related published works to validate the novelty of the work. RMSE, MAE, 9) Authors are claiming that the proposed hybrid method exhibits high level of performance compared to the isolated SVD implementations and allow much better interpretability than GNN based approaches. To validate such claim, it is suggested to include isolated SVD and GNN methods results in a single plot with that of the proposed method in the results and analysis section. 10) It is suggested to specify some limitations of the proposed method in the conclusion section so that the future scopes can be established. 11) Before submission of the revised version, it is suggested to check the minor typos and grammatical errors in the manuscript. Correct all such issues. Overall Decision: MAJOR REVISION. ********** what does this mean? ). If published, this will include your full peer review and any attached files.). If published, this will include your full peer review and any attached files.). If published, this will include your full peer review and any attached files.). If published, this will include your full peer review and any attached files. If you choose “no”, your identity will remain anonymous but your review may still be made public. Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our For information about this choice, including consent withdrawal, please see our For information about this choice, including consent withdrawal, please see our For information about this choice, including consent withdrawal, please see our Privacy Policy ..--> Reviewer #1: Yes: Jawad K. AliJawad K. AliJawad K. AliJawad K. Ali Reviewer #2: No Reviewer #3: No Reviewer #4: Yes: DR. TARUN KUMAR DASDR. TARUN KUMAR DASDR. TARUN KUMAR DASDR. TARUN KUMAR DAS ********** [NOTE: If reviewer comments were submitted as an attachment file, they will be attached to this email and accessible via the submission site. Please log into your account, locate the manuscript record, and check for the action link "View Attachments". If this link does not appear, there are no attachment files.] To ensure your figures meet our technical requirements, please review our figure guidelines: https://journals.plos.org/plosone/s/figures You may also use PLOS’s free figure tool, NAAS, to help you prepare publication quality figures: https://journals.plos.org/plosone/s/figures#loc-tools-for-figure-preparation . NAAS will assess whether your figures meet our technical requirements by comparing each figure against our figure specifications. Attachment Submitted filename: PONE-D-25-63400_suggestions and comments.pdf pone.0346579.s001.pdf (1.2MB, pdf) PLoS One. 2026 Apr 13;21(4):e0346579. doi: 10.1371/journal.pone.0346579.r002 Author response to Decision Letter 1 Article notes Copyright and License information Collection date 2026. PMC Copyright notice 3 Feb 2026 Dear Academic Editor and Reviewers, We would like to thank you for the opportunity to revise our manuscript. We appreciate the thoughtful and constructive comments provided by the reviewers, which have significantly helped us improve the quality, clarity, and technical rigor of our work. We have addressed each comment point-by-point below and have highlighted the corresponding changes in the revised manuscript. Major revisions include the addition of a comparative analysis against GNNs, a rigorous complexity analysis, hyperparameter tuning details, and a complete restructuring of the text into the passive voice as requested. 1. The PLOS LaTeX template has been followed using Overleaf Latex software. 2. The “minimal data set” has been submitted as additional information for your kind reference. Response to Reviewer #1 Comment 1-4 & 6: The reviewer noted that the article aligns well with the journal's scope, is organized, and clearly presents the methodology. Response: We thank Reviewer #1 for their encouraging comments and for recognizing the alignment of our work with the journal's scope. We appreciate the positive feedback regarding the clarity of our presentation. Comment 5: The authors concluded that these metrics demonstrate high performance compared to isolated SVD implementations and offer significantly better interpretability than GNN-based approaches. Response: We appreciate this observation. To substantiate this conclusion with concrete evidence, we have added a new "Comparative Analysis and Discussion" subsection in the Results section. • Table 3 has been added to explicitly benchmark our model’s RMSE (0.9966) against isolated SVD (typically ~1.05 on sparse data) and GNN-based approaches (typically ~0.90-0.95). • We have expanded the discussion to highlight that while GNNs offer a slight accuracy edge, our hybrid method offers superior interpretability through community structures, validating the trade-off claimed in our conclusion. Response to Reviewer #2 Comment 1: The abstract makes strong claims... but does not clearly explain the novelty... Adding a clearer statement of the methodological contribution would strengthen the abstract. Response: We have rewritten the Abstract to explicitly define the "sparsity-interpretability trade-off" as the core problem. We clarified that our novelty lies in using Louvain modularity not just for regularization, but as a pre-processing step to create dense, homogenous sub-matrices for the SVD, thereby reducing noise. Comment 2: The description of experimental results lacks context—no baseline RMSE/MAE values... Response: We have added a "Comparative Analysis" subsection in the Results section. This includes Table 3, which provides standard literature benchmarks for Standard User-CF, Basic SVD, and Neural CF (Deep Learning), allowing readers to contextualize our RMSE of 0.9966. Comment 3: The introduction should clearly conclude with a distinct section highlighting the novel contributions. Response: We have added a bulleted "Contributions" subsection at the end of the Introduction to explicitly list the three key contributions of this study: the novel hybrid architecture, enhanced interpretability, and scalable performance. Comment 4: Add a para that mentions briefly what each next section contains. Response: We have added an "Organization of the Paper" subsection at the end of the Introduction that outlines the structure of the subsequent sections. Comment 5: The literature review should benefit from more explorations of previous studies. Response: We have expanded the Related Work section to include a discussion on recent studies involving side-information and lightweight community detection methods, positioning our work within this broader context. Comment 6: The discussion section needs to be expanded to more thoroughly analyze the results. Response: We have expanded the Results and Discussion section to explain why the proposed method performs well. Specifically, we discuss how the "divide-and-conquer" strategy isolates noise within specific communities, leading to more precise latent feature extraction. Comment 7: The first paragraph of the conclusion should succinctly summarize the contributions of the study in past tense. Response: We have rewritten the first paragraph of the Conclusion in the past tense to summarize exactly what was done and what was achieved. Comment 8: The second paragraph of the conclusion should provide clear and actionable future recommendations. Response: We have revised the future work paragraph to be specific and actionable, suggesting the incorporation of temporal dynamics and testing on distributed computing frameworks. Comment 9: Equations are not properly cited. Response: We have added the appropriate citations (e.g., [Soman et al.], [Blondel et al.]) immediately preceding the mathematical formulations for SVD and Modularity in the Methodology section. Comment 10: Manuscript is highly low on visualizations. Response: To address this, we have added Figure 6, a methodological flowchart that visually describes the step-by-step architecture of the proposed system. We have also added Figure 7, a bar chart comparing our RMSE against baselines. Response to Reviewer #3 Comment 1: The introduction can include more recent state of the art works. Response: We have updated the Introduction to reference recent trends, including Transformer-based models and Large Language Models (LLMs) for sequential recommendation, to provide a more current state-of-the-art context. Comment 2: Please specify the novelty and contributions of the work. Response: As noted in our response to Reviewer #2, we have added a specific "Contributions" subsection to the Introduction. Comment 3: The assumptions... along with possible identifiability SVD conditions should be mentioned. Response: We have added a new subsection titled "Assumptions and Theoretical Constraints" in the Methodology section. This explicitly outlines our assumptions regarding the Low-Rank nature of the matrix, Homophily, and SVD Identifiability. Comment 4 & 6: It is not clear how to choose the learning rate and regularization parameter... Simulations should show the impact. Response: We have added a "Hyperparameter Tuning" subsection in the Proposed Method section. This details our Grid Search strategy and reports the specific values selected ($\alpha = 0.005$, $\lambda = 0.02$) and the observed effects of varying these parameters. Comment 5: It is not clear how diversity is treated. Response: We have added a "Discussion on Diversity and Personalization" in the Results section. We explain that while our primary metric is RMSE, the community-based approach inherently improves diversity by mitigating popularity bias and surfacing niche items relevant to specific user groups. Comment 7: The authors should conduct a rigorous computational complexity analysis. Response: We have added a "Computational Complexity Analysis" subsection in the Methodology. We use Big-O notation to demonstrate that our approach (O(N log N) for clustering + Local SVD) is computationally more efficient than standard GNN-based methods (O(L . |R| + L . N . d^2)). Comment 8: The work should compare with more state of the art methods. Response: We have addressed this via Table 3 and the new Figure 7, which compare our method against GNN-based benchmarks (like NGCF) and standard SVD. Comment 9: Please fix the text size of the figure, for example Fig. 1 has very small text. Response: We have revised Figure 1 to ensure that the axis labels and text are legible and of appropriate size for publication. Response to Reviewer #4 Comment 1: Avoid the word “we”, “our”... Write in passive voice. Response: We have meticulously revised the entire manuscript (Abstract, Introduction, Conclusion, and Methodology) to replace active voice (e.g., "We propose") with passive voice (e.g., "A method is proposed"). Comment 2: The citations in the Introduction have referred randomly... Rearrange the reference list. Response: We have completely reordered the Bibliography and citations throughout the text. References now appear sequentially, starting with [1], [2] in the Introduction. Comment 3: Write concluding remarks at the end of the related work about what new thing has been focused on... Response: We have added a subsection "Concluding Remarks on Research Gap" at the end of Related Work to clearly articulate the gap our study fills regarding the trade-off between sparsity and interpretability. Comment 4: Include a flowchart at the beginning of section 4. Response: We have inserted Figure 6 (Architectural Workflow) at the beginning of the Proposed Method section (Section 4) to provide an immediate visual overview of the system design. Comment 5: Figure 1 is not mentioned in the text. Response: We have added an explicit reference to Figure 1 in the Results section, using it to discuss the long-tail distribution of the dataset ratings. Comment 6: Elaborate Figure 2, 3 and 4 in more details. Response: We have added detailed descriptive text in the Results section interpreting specific features of these figures (e.g., interpreting the connectivity density in the network visualization and the user distribution across clusters). Comment 7 & 9: Results section needs improvement... include isolated SVD and GNN methods results in a single plot. Response: We have added Figure 7, a comparative bar chart that visually displays the RMSE of Isolated SVD, GNN, and the Proposed Hybrid method side-by-side, validating our performance claims. Comment 8: Include a comparative table. Response: We have included Table 3 in the Results section, which compares Methodology, RMSE, Sparsity Handling, and Interpretability across the different approaches. Comment 10: Specify some limitations of the proposed method in the conclusion. Response: We have added a paragraph to the Conclusion explicitly stating the limitations of our method, specifically regarding the "cold-start" issue for new users and the computational cost of periodic re-clustering. Comment 11: Check minor typos and grammatical errors. Response: We have proofread the manuscript and corrected typos and grammatical errors to ensure a formal academic tone. Attachment Submitted filename: Response to Reviewers.docx pone.0346579.s002.docx (19.8KB, docx) PLoS One. doi: 10.1371/journal.pone.0346579.r003 Decision Letter 1 Shih-Lin Lin Shih-Lin Lin Academic Editor Find articles by Shih-Lin Lin Author information Copyright and License information Roles Shih-Lin Lin : Academic Editor © 2026 Shih-Lin Lin This is an open access article distributed under the terms of the Creative Commons Attribution License , which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. PMC Copyright notice 15 Feb 2026 Dear Dr. Rajathi G, Thank you for submitting your manuscript to PLOS ONE. After careful consideration, we feel that it has merit but does not fully meet PLOS ONE’s publication criteria as it currently stands. Therefore, we invite you to submit a revised version of the manuscript that addresses the points raised during the review process. Please submit your revised manuscript by Apr 01 2026 11:59PM. If you will need more time than this to complete your revisions, please reply to this message or contact the journal office at [email protected] . When you're ready to submit your revision, log on to https://www.editorialmanager.com/pone/ and select the 'Submissions Needing Revision' folder to locate your manuscript file.. When you're ready to submit your revision, log on to https://www.editorialmanager.com/pone/ and select the 'Submissions Needing Revision' folder to locate your manuscript file.. When you're ready to submit your revision, log on to https://www.editorialmanager.com/pone/ and select the 'Submissions Needing Revision' folder to locate your manuscript file.. When you're ready to submit your revision, log on to https://www.editorialmanager.com/pone/ and select the 'Submissions Needing Revision' folder to locate your manuscript file. A letter that responds to each point raised by the academic editor and reviewer(s). You should upload this letter as a separate file labeled 'Response to Reviewers'. A marked-up copy of your manuscript that highlights changes made to the original version. You should upload this as a separate file labeled 'Revised Manuscript with Track Changes'. An unmarked version of your revised paper without tracked changes. You should upload this as a separate file labeled 'Manuscript'. If applicable, we recommend that you deposit your laboratory protocols in protocols.io to enhance the reproducibility of your results. Protocols.io assigns your protocol its own identifier (DOI) so that it can be cited independently in the future. For instructions see: https://journals.plos.org/plosone/s/submission-guidelines#loc-laboratory-protocols . Additionally, PLOS ONE offers an option for publishing peer-reviewed Lab Protocol articles, which describe protocols hosted on protocols.io. Read more information on sharing protocols at . Additionally, PLOS ONE offers an option for publishing peer-reviewed Lab Protocol articles, which describe protocols hosted on protocols.io. Read more information on sharing protocols at . Additionally, PLOS ONE offers an option for publishing peer-reviewed Lab Protocol articles, which describe protocols hosted on protocols.io. Read more information on sharing protocols at . Additionally, PLOS ONE offers an option for publishing peer-reviewed Lab Protocol articles, which describe protocols hosted on protocols.io. Read more information on sharing protocols at https://plos.org/protocols?utm_medium=editorial-email&utm_source=authorletters&utm_campaign=protocols .... We look forward to receiving your revised manuscript. Kind regards, Shih-Lin Lin, Ph.D Academic Editor PLOS One Journal Requirements: 1. If the reviewer comments include a recommendation to cite specific previously published works, please review and evaluate these publications to determine whether they are relevant and should be cited. There is no requirement to cite these works unless the editor has indicated otherwise. 2. Please review your reference list to ensure that it is complete and correct. If you have cited papers that have been retracted, please include the rationale for doing so in the manuscript text, or remove these references and replace them with relevant current references. Any changes to the reference list should be mentioned in the rebuttal letter that accompanies your revised manuscript. If you need to cite a retracted article, indicate the article’s retracted status in the References list and also include a citation and full reference for the retraction notice. [Note: HTML markup is below. Please do not edit.] Reviewers' comments: Reviewer's Responses to Questions Comments to the Author Reviewer #2: All comments have been addressed Reviewer #3: (No Response) Reviewer #4: All comments have been addressed ********** 2. Is the manuscript technically sound, and do the data support the conclusions??> Reviewer #2: (No Response) Reviewer #3: (No Response) Reviewer #4: Yes ********** 3. Has the statistical analysis been performed appropriately and rigorously? -->?> Reviewer #2: (No Response) Reviewer #3: (No Response) Reviewer #4: Yes ********** 4. Have the authors made all data underlying the findings in their manuscript fully available??> The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.--> Reviewer #2: Yes Reviewer #3: (No Response) Reviewer #4: Yes ********** 5. Is the manuscript presented in an intelligible fashion and written in standard English??> Reviewer #2: (No Response) Reviewer #3: (No Response) Reviewer #4: Yes ********** Reviewer #2: The responses are generally good but the manuscript has a few typos that the authors should address. Reviewer #3: Most of my previous comments have been addressed. Please revise the following ones: 1. General comment: Although the authors argue that this sequential approach improves interpretability and reduces computational memory requirements compared to GNN-based methods, the empirical results on the Netflix Prize dataset reveal that the proposed hybrid model yields a higher RMSE (0.9966) than the state-of-the-art baselines (0.92). Please explain. 2. There are some grammatical issues like “the matrix is not sparse however, it is connected enough inside…”. Please revise. 3. Double check format of references for consistency. Reviewer #4: I am pleased to inform that the paper is now ready for publication. Congratulations for this good work. ********** what does this mean? ). If published, this will include your full peer review and any attached files.). If published, this will include your full peer review and any attached files.). If published, this will include your full peer review and any attached files.). If published, this will include your full peer review and any attached files. If you choose “no”, your identity will remain anonymous but your review may still be made public. Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our For information about this choice, including consent withdrawal, please see our For information about this choice, including consent withdrawal, please see our For information about this choice, including consent withdrawal, please see our Privacy Policy ..--> Reviewer #2: No Reviewer #3: No Reviewer #4: Yes: DR. TARUN KUMAR DASDR. TARUN KUMAR DASDR. TARUN KUMAR DASDR. TARUN KUMAR DAS ********** [NOTE: If reviewer comments were submitted as an attachment file, they will be attached to this email and accessible via the submission site. Please log into your account, locate the manuscript record, and check for the action link "View Attachments". If this link does not appear, there are no attachment files.] To ensure your figures meet our technical requirements, please review our figure guidelines: https://journals.plos.org/plosone/s/figures You may also use PLOS’s free figure tool, NAAS, to help you prepare publication quality figures: https://journals.plos.org/plosone/s/figures#loc-tools-for-figure-preparation . NAAS will assess whether your figures meet our technical requirements by comparing each figure against our figure specifications. PLoS One. 2026 Apr 13;21(4):e0346579. doi: 10.1371/journal.pone.0346579.r004 Author response to Decision Letter 2 Article notes Copyright and License information Collection date 2026. PMC Copyright notice 16 Mar 2026 Title: Enhancing Recommendation Systems through SVD-Based Collaborative Filtering and Community Detection Dear Editor and Reviewers, We sincerely thank the Editor and all Reviewers for their constructive feedback and positive assessments of our manuscript. The comments have been very helpful in improving the clarity, rigor, and presentation of our work. We have carefully addressed every point raised, with changes highlighted in the "Revised Manuscript with Track Changes" file. Below, we provide a point-by-point response to each comment. Most reviewers indicated that previous comments were fully addressed, and Reviewer #4 stated the paper is now ready for publication. We greatly appreciate this feedback. Reviewer #2 Comment: "The responses are generally good but the manuscript has a few typos that the authors should address." Response: Thank you for this observation. We have conducted a thorough proofread of the entire manuscript and corrected all identified typos and minor grammatical issues. Examples include: • "predictes" → "predicts" (Abstract) • "Hypersale" → "Hyperscale" (Contributions section) • Removed redundant "Mean value of 0.9966" phrase as it is mentioned in Abstract (kept only RMSE and MAE) • Fixed awkward phrasing in the first sentence of the Abstract ("low data density. scalability...") → "low data density, scalability issues, and lack of interpretability." • Other minor fixes (e.g., sentence flow in Abstract conclusion: "paves the path... appropriate to large-scale sparse. datasets" → "enables scalable and transparent recommendation engines suitable for large-scale sparse datasets.") These changes improve readability and polish. All edits are visible in the tracked-changes version (primarily pages 1–2 and scattered minor spots). Reviewer #3 Comment 1: "Although the authors argue that this sequential approach improves interpretability and reduces computational memory requirements compared to GNN-based methods, the empirical results on the Netflix Prize dataset reveal that the proposed hybrid model yields a higher RMSE (0.9966) than the state-of-the-art baselines (0.92). Please explain." Response: Thank you for highlighting this important point. We have added explicit explanations of the RMSE difference to emphasize the deliberate trade-off in our method. • In the Abstract (final paragraph, after reporting RMSE/MAE): Added phrasing to note that the model achieves "competitive predictive performance with significantly higher interpretability and lower computational cost than complex deep learning baselines, despite a modestly higher RMSE due to the deliberate trade-off for transparency and efficiency on extremely sparse data." • In the Results and Discussion → Comparative Analysis subsection (immediately after mentioning the ~0.92 for GNNs): Added sentences explaining: "This difference arises because the proposed method prioritizes interpretability (clear community-based explanations) and computational efficiency (local factorization on denser sub-matrices) over marginal accuracy gains from black-box GNN architectures, which require substantially higher resources on sparse datasets like Netflix." These additions clarify that the modest RMSE increase is acceptable given the gains in interpretability and scalability. Changes appear on approximately pages 1 (Abstract) and 12 (Discussion) in the tracked manuscript. Comment 2: "There are some grammatical issues like “the matrix is not sparse however, it is connected enough inside…”. Please revise." Response: We have revised the specific sentence in the Assumptions and Theoretical Constraints subsection (point 3 of the enumerated list) for clarity and proper punctuation: Original: "Identifiability with SVD: (To achieve the unique reconstruction of a rating matrix, up to a permutation), one assumes that the singular values are not identical and that the matrix is not sparse however, it is connected enough inside the community to converge during Stochastic Gradient Descent (SGD)." Revised: "Identifiability with SVD: To achieve unique reconstruction of the rating matrix (up to permutation), one assumes that the singular values are distinct and that intra-community sub-matrices, while not fully dense, are sufficiently connected to ensure convergence during Stochastic Gradient Descent (SGD)." This fixes the run-on structure and improves readability. Change is on approximately page 5. Comment 3: "Double check format of references for consistency." Response: Thank you for this suggestion. We have switched to BibTeX processing using the PLOS-provided plos2025.bst style file and a standardized refs.bib file. This ensures uniform formatting across all entries (e.g., consistent author initials, volume/page/year placement, DOI inclusion where available, and proper handling of arXiv/journal types). No retracted articles are cited. All changes are reflected in the revised bibliography section (approximately page 15 onward). We believe this fully addresses consistency concerns. Reviewer #4 Comment: "I am pleased to inform that the paper is now ready for publication. Congratulations for this good work." Response: We are very grateful for this positive evaluation and for Reviewer #4's kind words. No further changes were required based on this review. In addition to the above, we performed a final global proofread for minor flow and consistency (e.g., variable naming in equations: sim(x, y) → sim(u, v) to match usage; table caption updated from "Singular Value Decomposition on test set" → "Proposed Hybrid Model on test set" for accuracy). No new data, analyses, or major rewrites were introduced. Few more appending were as follows: Data Availability: The dataset used in this study is the publicly available Netflix Prize dataset, which contains anonymized movie rating interactions between users and movies and is widely used for benchmarking recommender systems. The dataset includes millions of user–item ratings collected during the Netflix Prize competition and is designed for research in collaborative filtering and recommendation algorithms. The dataset can be accessed from the Kaggle repository at: { https://www.kaggle.com/datasets/netflix-inc/netflix-prize-data }. The dataset from the original is reproduced in the following dataset doi:10.6084/m9.figshare.31428857 Funding: This research received no external funding. The work was conducted as part of academic research only. We believe these revisions fully address all concerns and significantly strengthen the manuscript. Thank you again for your time and consideration. We look forward to your decision. Sincerely, Ignisha Rajathi G (Corresponding Author) Manipal Institute of Technology Bengaluru, Manipal Academy of Higher Education Email: [email protected] On behalf of all authors: Keerthika T, Vedhapriyavadhana Rajamani, Surya Santhosh Kumar, Karthigai Selvam, Aiyyappan M R, L S Thoshi Babu ________________________________________ Attachment Submitted filename: Response to Reviewers-2.docx pone.0346579.s003.docx (15.1KB, docx) PLoS One. doi: 10.1371/journal.pone.0346579.r005 Decision Letter 2 Shih-Lin Lin Shih-Lin Lin Academic Editor Find articles by Shih-Lin Lin Author information Copyright and License information Roles Shih-Lin Lin : Academic Editor © 2026 Shih-Lin Lin This is an open access article distributed under the terms of the Creative Commons Attribution License , which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. PMC Copyright notice 22 Mar 2026 Enhancing Recommendation Systems through SVD-Based Collaborative Filtering and Community Detection PONE-D-25-63400R2 Dear Dr. Rajathi G, We’re pleased to inform you that your manuscript has been judged scientifically suitable for publication and will be formally accepted for publication once it meets all outstanding technical requirements. No further revisions are required before publication. In addition, the additional citation suggested by Reviewer 3 is not required for acceptance. Within one week, you’ll receive an e-mail detailing the required amendments. When these have been addressed, you’ll receive a formal acceptance letter and your manuscript will be scheduled for publication. An invoice will be generated when your article is formally accepted. Please note, if your institution has a publishing partnership with PLOS and your article meets the relevant criteria, all or part of your publication costs will be covered. Please make sure your user information is up-to-date by logging into Editorial Manager at Editorial Manager® and clicking the ‘Update My Information' link at the top of the page. For questions related to billing, please contact  and clicking the ‘Update My Information' link at the top of the page. For questions related to billing, please contact  and clicking the ‘Update My Information' link at the top of the page. For questions related to billing, please contact  and clicking the ‘Update My Information' link at the top of the page. For questions related to billing, please contact billing support .... If your institution or institutions have a press office, please notify them about your upcoming paper to help maximize its impact. If they’ll be preparing press materials, please inform our press team as soon as possible -- no later than 48 hours after receiving the formal acceptance. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information, please contact [email protected]. Kind regards, Shih-Lin Lin, Ph.D Academic Editor PLOS One Additional Editor Comments (optional): Reviewers' comments: Reviewer's Responses to Questions Comments to the Author Reviewer #2: (No Response) Reviewer #3: (No Response) ********** 2. Is the manuscript technically sound, and do the data support the conclusions??> Reviewer #2: (No Response) Reviewer #3: (No Response) ********** 3. Has the statistical analysis been performed appropriately and rigorously? -->?> Reviewer #2: (No Response) Reviewer #3: (No Response) ********** 4. Have the authors made all data underlying the findings in their manuscript fully available??> The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.--> Reviewer #2: (No Response) Reviewer #3: (No Response) ********** 5. Is the manuscript presented in an intelligible fashion and written in standard English??> Reviewer #2: (No Response) Reviewer #3: (No Response) ********** Reviewer #2: (No Response) Reviewer #3: I have the following minor comments: 1. The following works on deep learning can be included to enrich your introduction https://doi.org/10.1109/VTC2023-Spring57618.2023.10200157 and https://doi.org/10.1109/GCWkshps58843.2023.10464930 . ********** what does this mean? ). If published, this will include your full peer review and any attached files.). If published, this will include your full peer review and any attached files.). If published, this will include your full peer review and any attached files.). If published, this will include your full peer review and any attached files. If you choose “no”, your identity will remain anonymous but your review may still be made public. Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our For information about this choice, including consent withdrawal, please see our For information about this choice, including consent withdrawal, please see our For information about this choice, including consent withdrawal, please see our Privacy Policy ..--> Reviewer #2: No Reviewer #3: No ********** PLoS One. doi: 10.1371/journal.pone.0346579.r006 Acceptance letter Shih-Lin Lin Shih-Lin Lin Academic Editor Find articles by Shih-Lin Lin Author information Copyright and License information Roles Shih-Lin Lin : Academic Editor © 2026 Shih-Lin Lin This is an open access article distributed under the terms of the Creative Commons Attribution License , which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. PMC Copyright notice PONE-D-25-63400R2 PLOS One Dear Dr. G, I'm pleased to inform you that your manuscript has been deemed suitable for publication in PLOS One. Congratulations! Your manuscript is now being handed over to our production team. At this stage, our production department will prepare your paper for publication. This includes ensuring the following: * All references, tables, and figures are properly cited * All relevant supporting information is included in the manuscript submission, * There are no issues that prevent the paper from being properly typeset You will receive further instructions from the production team, including instructions on how to review your proof when it is ready. Please keep in mind that we are working through a large volume of accepted articles, so please give us a few days to review your paper and let you know the next and final steps. Lastly, if your institution or institutions have a press office, please let them know about your upcoming paper now to help maximize its impact. If they'll be preparing press materials, please inform our press team within the next 48 hours. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information, please contact [email protected]. You will receive an invoice from PLOS for your publication fee after your manuscript has reached the completed accept phase. If you receive an email requesting payment before acceptance or for any other service, this may be a phishing scheme. Learn how to identify phishing emails and protect your accounts at https://explore.plos.org/phishing . If we can help with anything else, please email us at [email protected]. Thank you for submitting your work to PLOS ONE and supporting open access. Kind regards, PLOS ONE Editorial Office Staff on behalf of Professor Shih-Lin Lin Academic Editor PLOS One Associated Data This section collects any data citations, data availability statements, or supplementary materials included in this article. Supplementary Materials Attachment Submitted filename: PONE-D-25-63400_suggestions and comments.pdf pone.0346579.s001.pdf (1.2MB, pdf) Attachment Submitted filename: Response to Reviewers.docx pone.0346579.s002.docx (19.8KB, docx) Attachment Submitted filename: Response to Reviewers-2.docx pone.0346579.s003.docx (15.1KB, docx) Data Availability Statement All relevant data are within the paper. Further the dataset link is provided in the manuscript. The dataset can be accessed from the Kaggle repository at: { https://www.kaggle.com/datasets/netflix-inc/netflix-prize-data }. The dataset from the original is reproduced in the following dataset doi: 10.6084/m9.figshare.31428857 . Articles from PLOS One are provided here courtesy of PLOS ACTIONS View on publisher site PDF (1.1 MB) Cite Collections Permalink PERMALINK Copy RESOURCES Similar articles Cited by other articles Links to NCBI Databases Cite Copy Download .nbib .nbib Format: AMA APA MLA NLM Add to Collections Create a new collection Add to an existing collection Name your collection * Choose a collection Unable to load your collection due to an error Please try again Add Cancel Follow NCBI NCBI on X (formerly known as Twitter) NCBI on Facebook NCBI on LinkedIn NCBI on GitHub NCBI RSS feed Connect with NLM NLM on X (formerly known as Twitter) NLM on Facebook NLM on YouTube National Library of Medicine 8600 Rockville Pike Bethesda, MD 20894 Web Policies FOIA HHS Vulnerability Disclosure Help Accessibility Careers NLM NIH HHS USA.gov Back to Top