2014/05/08 by Emanuel Lacić, Lacic, Emanuel, Dominik Kowald +9 · 1 citation
Computer Science · Physics and Astronomy · Social Sciences · #Complex Network Analysis Techniques #Digital Marketing and Social Media #FOS: Computer and information sciences #Information Retrieval (cs.IR) #Recommender Systems and Techniques
paper · pdf · doi:10.48550/arxiv.1405.1837
openalex publication_date 2014/05/08 · openalex created_date 2025/10/24 · openalex updated_date 2026/07/28
Recent research has unveiled the importance of online social networks for\nimproving the quality of recommender systems and encouraged the research\ncommunity to investigate better ways of exploiting the social information for\nrecommendations. To contribute to this sparse field of research, in this paper\nwe exploit users' interactions along three data sources (marketplace, social\nnetwork and location-based) to assess their performance in a barely studied\ndomain: recommending products and domains of interests (i.e., product\ncategories) to people in an online marketplace environment. To that end we\ndefined sets of content- and network-based user similarity features for each\ndata source and studied them isolated using an user-based Collaborative\nFiltering (CF) approach and in combination via a hybrid recommender algorithm,\nto assess which one provides the best recommendation performance.\nInterestingly, in our experiments conducted on a rich dataset collected from\nSecondLife, a popular online virtual world, we found that recommenders relying\non user similarity features obtained from the social network data clearly\nyielded the best results in terms of accuracy in case of predicting products,\nwhereas the features obtained from the marketplace and location-based data\nsources also obtained very good results in case of predicting categories. This\nfinding indicates that all three types of data sources are important and should\nbe taken into account depending on the level of specialization of the\nrecommendation task.\n