Evaluation of Collaborative Filtering Algorithms for Recommending Articles on CiteULike

Workshop on Web 3.0: Merging Semantic Web and Social Web

Denis Parra ; Peter Brusilovsky

Others  :  http://CEUR-WS.org/Vol-467/paper5.pdf
PID  :  11157

来源: CEUR

【 摘 要 】

Motivated by the potential use of collaborative tagging systems todevelop new recommender systems, we have implemented andcompared three variants of user-based collaborative filteringalgorithms to provide recommendations of articles on CiteULike.On our first approach, Classic Collaborative filtering (CCF), weuse Pearson correlation to calculate similarity between users and aclassic adjusted ratings formula to rank the recommendations. Our second approach, Neighbor-weighted Collaborative Filtering(NwCF), incorporates the amount of raters in the ranking formulaof the recommendations. A modified version of the Okapi BM25IR model over users' tags is implemented on our third approach toform the user neighborhood. Our results suggest thatincorporating the number of raters into the algorithms leads to animprovement of precision, and they also support that tags can beconsidered as an alternative to Pearson correlation to calculate thesimilarity between users and their neighbors in a collaborativetagging system.