Collaborative filtering over distributed environment

Shlomo Berkovsky, Paolo Busetta, Yaniv Eytani, Tsvi Kuflik, Francesco Ricci

Research output: Chapter in Book/Report/Conference proceedingConference proceeding contributionpeer-review


Currently, implementations of the Collaborative Filtering (CF) algorithm are mostly centralized. Hence, information about the users, for example, product ratings, is concentrated in a single location. In this work we propose a novel approach to overcome the inherent limitations of CF (sparsity of data and cold start) by exploiting multiple distributed information repositories. These may belong to a single domain or to different domains. To facilitate our approach, we used LoudVoice, a multi-agent communication infrastructure that can connect similar information repositories into a single virtual structure called "implicit organization". Repositories are partitioned between such organizations according to geographical or topical criteria. We employ CF to generate user-personalized recommendations over different data distribution policies. Experimental results demonstrate that topical distribution outperforms geographical distribution. We also show that in geographical distribution using filtering based on social characteristics of the users improves the quality of recommendations.
Original languageEnglish
Title of host publicationDecentralized, Agent-Based and Social Approaches to User Modeling
Subtitle of host publicationWorkshop DASUM-05. Proceedings
EditorsPeter Dolog, Julita Vassileva
Number of pages10
Publication statusPublished - 2005
Externally publishedYes
EventInternational Conference on User Modeling (10th : 2005) - Edinburgh, United Kingdom
Duration: 24 Jul 200529 Jul 2005


ConferenceInternational Conference on User Modeling (10th : 2005)
Country/TerritoryUnited Kingdom


Dive into the research topics of 'Collaborative filtering over distributed environment'. Together they form a unique fingerprint.

Cite this