Design and Analysis of a Gossip-based Decentralized Trust Recommender System
2012 (English)In: 4th ACM Recommender Systems (RecSys) Workshop on Recommender Systems & the Social Web, 2012Conference paper, Presentation (Refereed)
Information overload has become an increasingly common problem in today’s large scale internet applications. Collaborative filtering(CF) recommendation systems have emerged as a popular solution to this problem by taking advantage of underlying social networks. Traditional CF recommenders suffer from lack of scalability while decentralized recommendation systems (DHT-based, Gossip-based etc.) have promised to alleviate this problem. Thus, in this paper we propose a decentralized approach to CF recommender sys tems that takes advantage of the popular P2P T-Man algorithm to create and maintain an overlay network capable of generating predictions based on only local information. We analyze our approaches performance in terms of prediction accuracy and item-coverage function of neighborhood size as well as number of T-Man rounds. We show our system achieves better accuracy than previous approaches while implementing a highly scalable, decentralized paradigm. We also show our system is able to generate predictions for a large fraction of users, which is comparable with the centralized approaches.
Place, publisher, year, edition, pages
Trust, Decentralized, Gossip, Recommender Systems
IdentifiersURN: urn:nbn:se:kth:diva-118489OAI: oai:DiVA.org:kth-118489DiVA: diva2:606541
Recommender Systems and the Social Web (RSWEB'2012)
QC 201302202013-02-192013-02-192014-01-24Bibliographically approved