Recommendation of users in social networks: A semantic and social based classification approach. Issue 2 (13th September 2020)

• Record Type: Journal Article
• Authors: Berkani, Lamia; Belkacem, Sami; Ouafi, Mounira; Guessoum, Ahmed
• Is Part Of: Expert systems. Volume 38:Issue 2(2021)
• Page Start: n/a
• Page End: n/a
• Publication Date: 2020-09-13
• Subjects: classification -- commitment -- credibility -- K‐means -- K‐Nearest Neighbours -- recommendation of users -- semantic filtering -- social filtering -- trust -- Yelp social network; Expert systems (Computer science); 006.33
• DOI: 10.1111/exsy.12634
• Languages: English
• ISSNs: 0266-4720

Abstract

Abstract: Recently, the study of social network‐based recommender systems has become an active research topic. The integration of the social relationships that exist between users can improve the accuracy of recommendation results since the users' preferences are similar or influenced by their connected friends. We focus in this article on the recommendation of users in social networks. Our approach is based on semantic and social representations of the users' profiles. We have formalized and illustrated these two dimensions using the Yelp social network. The novelty of our approach concerns the modelling of the credibility of the user, through his/her trust and commitment in the social network. Moreover, in order to optimize the performance of the recommendation process, we have used two classification techniques: an unsupervised technique that uses the K‐means algorithm (applied initially to all users); and a supervised technique that uses the K‐Nearest Neighbours algorithm (applied to newly added users). A recommendation algorithm has been proposed taking into account the cold‐start and sparsity problems. A prototype of a recommender system has been developed and tested using two publicly available datasets: the Yelp database and the Rich Epinions database. The comparative evaluation results show the effectiveness of combining the semantic, the social and the credibility information in an approach that appropriately uses the K‐means and K‐Nearest Neighbours algorithms.