Science and Technology Production
Congress
Authorship
CORBELLINI, A.
;
MATEOS, C.
;
GODOY, D.
;
ZUNINO, A.
;
SCHIAFFINO, S.
Date
2013
Publishing House and Editing Place
Brazilian Computing Society
Summary
Information provided by the agent in
SIGEVA
Most recommendation algorithms in the context of large-scale social networks such as Twitter or Facebook struggle with the need of an efficient exploration of the huge and exponentially growing user graph. Current solutions in the form of graph-specific databases or frameworks for graph algorithms do not scale well for processing complex navigational patterns. In this paper we present an approach for supporting social recommendation algorithms that operate with large graphs in a computer cluste...
Most recommendation algorithms in the context of large-scale social networks such as Twitter or Facebook struggle with the need of an efficient exploration of the huge and exponentially growing user graph. Current solutions in the form of graph-specific databases or frameworks for graph algorithms do not scale well for processing complex navigational patterns. In this paper we present an approach for supporting social recommendation algorithms that operate with large graphs in a computer cluster based on “policies”, rules that allow users to throttle the amount of parallelism and control task location. Experiments with a followee recommendation algorithm show the potentials of the proposed policies to solve recommendation problems in an efficient and scalable way.
Show more
Show less
Key Words
Social Recommender SystemsBig-dataDistributed SystemsSocial Networks