Investigating algorithmic variations of an RS Graph-based collaborative filtering approach

A personalized recommendation system learns users specific profiles from users feedback and content in order to deliver information tailored to each individual user's interest. Although great effort has been devoted to the proposal, implementation and study of recommendation systems approaches, there is still a lot of room for improvement. Currently, much research on recommender systems focuses on improving the accuracy of their algorithms. In a recent work, we proposed a novel Graph-based Collaborative Filtering approach for recommendation systems based on five steps: (1)-Creating a Homophily network using similarity measures, (2)- Identifying communities in the Homophily network, (3)- Identifying key nodes per community, (4)- Profiling key nodes per community and (5)-Computing recommendations for community users based on the resulting profiles. This paper delves deeper into the first two steps of our approach. In fact, the wide range of algorithms for community detection compelled us to create variations of our approach and to conduct a comparative analysis of their accuracy and time of execution. Results from testing our variations on two open datasets present comparable results. While the Louvain algorithm has the merit of simplicity and straightforwardness, AHC and K-means algorithms render better results when dealing with a larger datasets.

[1]  Kenneth C. Gehrt,et al.  A Shopping Orientation Segmentation of French Consumers: Implications for Catalog Marketing , 1998 .

[2]  Jonathan L. Herlocker,et al.  Evaluating collaborative filtering recommender systems , 2004, TOIS.

[3]  Maciej Kula,et al.  Metadata Embeddings for User and Item Cold-start Recommendations , 2015, CBRecSys@RecSys.

[4]  Jean-Loup Guillaume,et al.  Fast unfolding of communities in large networks , 2008, 0803.0476.

[5]  Yi Zhang,et al.  Interaction and Personalization of Criteria in Recommender Systems , 2010, UMAP.

[6]  G. W. Stewart,et al.  On the Early History of the Singular Value Decomposition , 1993, SIAM Rev..

[7]  F. Maxwell Harper,et al.  The MovieLens Datasets: History and Context , 2016, TIIS.

[8]  Mohammed Amine Janati Idrissi,et al.  Towards a Novel Graph-based collaborative filtering approach for recommendation systems , 2018, SITA.

[9]  Pranav Nerurkar,et al.  A Comparative Analysis of Community Detection Algorithms on Social Networks , 2018, Computational Intelligence: Theories, Applications and Future Directions - Volume I.

[10]  Trupti M. Kodinariya,et al.  Review on determining number of Cluster in K-Means Clustering , 2013 .

[11]  Siddique Latif,et al.  Community detection in networks: A multidisciplinary review , 2018, J. Netw. Comput. Appl..

[12]  Ayse Cufoglu,et al.  User Profiling - A Short Review , 2014 .

[13]  J. H. Ward.  Hierarchical Grouping to Optimize an Objective Function , 1963 .

[14]  K. Sasirekha,et al.  Agglomerative Hierarchical Clustering Algorithm- A Review , 2013 .

[15]  Sandra Sudarsky,et al.  Massive Quasi-Clique Detection , 2002, LATIN.

[16]  Robert Tibshirani,et al.  The Elements of Statistical Learning: Data Mining, Inference, and Prediction, 2nd Edition , 2001, Springer Series in Statistics.

[17]  Charu C. Aggarwal,et al.  Managing and Mining Graph Data , 2010, Managing and Mining Graph Data.