《大数据分析与处理》课程资料.pptxVIP

Recommender Systems:Latent Factor ModelsNote to other teachers and users of these slides: We would be delighted if you found this our material useful in giving your own lectures. Feel free to use these slides verbatim, or to modify them to fit your own needs. If you make use of a significant portion of these slides in your own lecture, please include this message, or a link to our web site: /info/1101/1266.htm The Netflix Prize?2 The Netflix Utility Matrix R1343554553 32225 2113 31480,000 users17,700 movies3Matrix R Utility Matrix R: Evaluation1343554553 32??? 21?3 ?1Test Data Set?4480,000 users17,700 moviesPredicted ratingTrue rating of user x on item i?Matrix RTraining Data Set BellKor Recommender SystemThe winner of the Netflix Challenge!Multi-scale modeling of the data:Combine top level, “regional”modeling of the data, with a refined, local view:Global:Overall deviations of users/moviesFactorization: Addressing “regional” effectsCollaborative filtering: Extract local patterns5Global effectsFactorizationCollaborative filtering Modeling Local Global EffectsGlobal:Mean movie rating: 3.7 starsThe Sixth Sense is 0.5 stars above avg.Joe rates 0.2 stars below avg. ? Baseline estimation: Joe will rate The Sixth Sense 4 starsLocal neighborhood (CF/NN):Joe didn’t like related movie Signs? Final estimate:Joe will rate The Sixth Sense 3.8 stars6 Recap: Collaborative Filtering (CF)Earliest and most popular collaborative filtering methodDerive unknown ratings from those of “similar” movies (item-item variant)Define similarity measure sij of items i and jSelect k-nearest neighbors, compute the rating N(i; x): items most similar to i that were rated by x7sij… similarity of items i and jrxj…rating of user x on item jN(i;x)… set of items similar to item i that were rated by x Modeling Local Global EffectsIn practice we get better estimates if we model deviations:8μ = overall mean ratingbx = rating deviation of user x = (avg. rating of user x) – μbi =