A user-item relevance model for log-based collaborative filtering

J Wang, AP de Vries, MJT Reinders

Research output: Chapter in Book/Conference proceedings/Edited volumeConference contributionScientificpeer-review

67 Citations (Scopus)

Abstract

.Implicitacquisitionofuserpreferencesmakeslog-basedcollaborative¿lteringfavorableinpracticetoaccomplishrecommendations.Inthispaper,wefollowaformalapproachintextretrievaltore-formulatetheproblem.Basedontheclassicprobabilityrankingprinciple,weproposeaprobabilisticuser-itemrelevancemodel.Underthisformalmodel,weshowthatuser-basedanditem-basedapproachesareonlytwodi¿erentfactorizationswithdi¿erentindependenceassumptions.Moreover,weshowthatsmoothingisanimportantaspecttoestimatetheparametersofthemodelsduetodatasparsity.Byaddinglinearinterpolationsmoothing,theproposedmodelgivesaprobabilisticjusti¿cationofusingTF×IDF-likeitemrankingincollaborative¿ltering.Besidesgivingtheinsightunderstandingoftheproblemofcollaborative¿ltering,wealsoshowexperimentsinwhichtheproposedmethodprovidesabetterrecommendationperformanceonamusicplay-listdataset.
Original languageUndefined/Unknown
Title of host publicationAdvances in information retrieval
EditorsM Lalmas, A MacFarlane, S Rüger, A Tombros, T Tsikrika, A Yavlinsky
Place of PublicationHeidelberg
PublisherSpringer
Pages37-48
Number of pages12
Publication statusPublished - 2006
Event28th European conference on IR research, ECIR 2006, London, UK - Heidelberg
Duration: 10 Apr 200612 Apr 2006

Publication series

Name
PublisherSpringer
NameLecture Notes in Computer Science
Volume3936
ISSN (Print)0302-9743

Conference

Conference28th European conference on IR research, ECIR 2006, London, UK
Period10/04/0612/04/06

Keywords

  • Wiskunde en Informatica
  • Techniek
  • technische Wiskunde en Informatica
  • conference contrib. refereed
  • CWTS 0.75 <= JFIS < 2.00

Cite this