DSpace Repository

Recommender System Optimization

Show simple item record

dc.contributor.advisor Dailey, Matthew N.
dc.contributor.author Yang, Jiawei
dc.contributor.other Esichaikul, Vatcharaporn
dc.contributor.other Guha, Sumanta
dc.date.accessioned 2016-07-21T06:42:43Z
dc.date.available 2016-07-21T06:42:43Z
dc.date.issued 2016-07
dc.identifier.uri http://www.cs.ait.ac.th/xmlui/handle/123456789/827
dc.description.abstract Collaborative filtering algorithms are among the most widely used in recommender systems. Since CF algorithms follow a common assumption, that similar users will give similar rating to similar items, I explore four techniques for improving traditional collaborative filtering algorithms, beginning with the Pointwise method. My first technique combines distance and ranking in similarity calculation. The second technique introduces a grade transfer factor between users. The third idea extends common items to similar item pairs supported by an assumption that similar users have similar interests based on similar products, whereas tradi-tional CF algorithms’ common assumption is that similar users have similar interests based on common products. The fourth technique predicts prediction error instead of just focus-ing on improving the prediction equation. MovieLense-1M is used as the test data set. The best results come from combining techniques 1 and 2, which provide a RMSE of 0.9478355, and from technique 3, with RMSE of 0.97649693, compared a RMSE of 1.050713 from the original Pointwie method I optimize.The results show that technique 1, 2 and 3 can improve NDCG compared with Poinwise, Pairwise, and Listwise. en_US
dc.description.sponsorship China Scholarship Council en_US
dc.language.iso en en_US
dc.publisher AIT en_US
dc.subject Recommender System, Collaborative Filtering en_US
dc.title Recommender System Optimization en_US
dc.type Research report en_US

Files in this item

This item appears in the following Collection(s)

Show simple item record

Search DSpace

Advanced Search


My Account