Recommending new movies: Even a few ratings are more valuable than metadata

István Pilászy, D. Tikk

Research output: Chapter in Book/Report/Conference proceedingConference contribution

88 Citations (Scopus)

Abstract

The Netflix Prize (NP) competition gave much attention to collaborative filtering (CF) approaches. Matrix factorization (MF) based CF approaches assign low dimensional feature vectors to users and items. We link CF and content-based filtering (CBF) by finding a linear transformation that transforms user or item descriptions so that they are as close as possible to the feature vectors generated by MF for CF. We propose methods for explicit feedback that are able to handle 140,000 features when feature vectors are very sparse. With movie metadata collected for the NP movies we show that the prediction performance of the methods is comparable to that of CF, and can be used to predict user preferences on new movies. We also investigate the value of movie metadata compared to movie ratings in regards of predictive power. We compare our solely CBF approach with a simple baseline rating-based predictor. We show that even 10 ratings of a new movie are more valuable than its metadata for predicting user ratings.

Original languageEnglish
Title of host publicationRecSys'09 - Proceedings of the 3rd ACM Conference on Recommender Systems
Pages93-100
Number of pages8
DOIs
Publication statusPublished - 2009
Event3rd ACM Conference on Recommender Systems, RecSys'09 - New York, NY, United States
Duration: Oct 23 2009Oct 25 2009

Other

Other3rd ACM Conference on Recommender Systems, RecSys'09
CountryUnited States
CityNew York, NY
Period10/23/0910/25/09

Fingerprint

Collaborative filtering
Metadata
Factorization
Linear transformations
Feedback

Keywords

  • Collaborative filtering
  • Content-based filtering
  • Matrix factorization
  • Netflix prize
  • NSVD1
  • RMSE

ASJC Scopus subject areas

  • Computer Science Applications
  • Software

Cite this

Pilászy, I., & Tikk, D. (2009). Recommending new movies: Even a few ratings are more valuable than metadata. In RecSys'09 - Proceedings of the 3rd ACM Conference on Recommender Systems (pp. 93-100) https://doi.org/10.1145/1639714.1639731

Recommending new movies : Even a few ratings are more valuable than metadata. / Pilászy, István; Tikk, D.

RecSys'09 - Proceedings of the 3rd ACM Conference on Recommender Systems. 2009. p. 93-100.

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Pilászy, I & Tikk, D 2009, Recommending new movies: Even a few ratings are more valuable than metadata. in RecSys'09 - Proceedings of the 3rd ACM Conference on Recommender Systems. pp. 93-100, 3rd ACM Conference on Recommender Systems, RecSys'09, New York, NY, United States, 10/23/09. https://doi.org/10.1145/1639714.1639731
Pilászy I, Tikk D. Recommending new movies: Even a few ratings are more valuable than metadata. In RecSys'09 - Proceedings of the 3rd ACM Conference on Recommender Systems. 2009. p. 93-100 https://doi.org/10.1145/1639714.1639731
Pilászy, István ; Tikk, D. / Recommending new movies : Even a few ratings are more valuable than metadata. RecSys'09 - Proceedings of the 3rd ACM Conference on Recommender Systems. 2009. pp. 93-100
@inproceedings{c6d6a253537a4f51b080f7ee80ea653c,
title = "Recommending new movies: Even a few ratings are more valuable than metadata",
abstract = "The Netflix Prize (NP) competition gave much attention to collaborative filtering (CF) approaches. Matrix factorization (MF) based CF approaches assign low dimensional feature vectors to users and items. We link CF and content-based filtering (CBF) by finding a linear transformation that transforms user or item descriptions so that they are as close as possible to the feature vectors generated by MF for CF. We propose methods for explicit feedback that are able to handle 140,000 features when feature vectors are very sparse. With movie metadata collected for the NP movies we show that the prediction performance of the methods is comparable to that of CF, and can be used to predict user preferences on new movies. We also investigate the value of movie metadata compared to movie ratings in regards of predictive power. We compare our solely CBF approach with a simple baseline rating-based predictor. We show that even 10 ratings of a new movie are more valuable than its metadata for predicting user ratings.",
keywords = "Collaborative filtering, Content-based filtering, Matrix factorization, Netflix prize, NSVD1, RMSE",
author = "Istv{\'a}n Pil{\'a}szy and D. Tikk",
year = "2009",
doi = "10.1145/1639714.1639731",
language = "English",
isbn = "9781605584355",
pages = "93--100",
booktitle = "RecSys'09 - Proceedings of the 3rd ACM Conference on Recommender Systems",

}

TY - GEN

T1 - Recommending new movies

T2 - Even a few ratings are more valuable than metadata

AU - Pilászy, István

AU - Tikk, D.

PY - 2009

Y1 - 2009

N2 - The Netflix Prize (NP) competition gave much attention to collaborative filtering (CF) approaches. Matrix factorization (MF) based CF approaches assign low dimensional feature vectors to users and items. We link CF and content-based filtering (CBF) by finding a linear transformation that transforms user or item descriptions so that they are as close as possible to the feature vectors generated by MF for CF. We propose methods for explicit feedback that are able to handle 140,000 features when feature vectors are very sparse. With movie metadata collected for the NP movies we show that the prediction performance of the methods is comparable to that of CF, and can be used to predict user preferences on new movies. We also investigate the value of movie metadata compared to movie ratings in regards of predictive power. We compare our solely CBF approach with a simple baseline rating-based predictor. We show that even 10 ratings of a new movie are more valuable than its metadata for predicting user ratings.

AB - The Netflix Prize (NP) competition gave much attention to collaborative filtering (CF) approaches. Matrix factorization (MF) based CF approaches assign low dimensional feature vectors to users and items. We link CF and content-based filtering (CBF) by finding a linear transformation that transforms user or item descriptions so that they are as close as possible to the feature vectors generated by MF for CF. We propose methods for explicit feedback that are able to handle 140,000 features when feature vectors are very sparse. With movie metadata collected for the NP movies we show that the prediction performance of the methods is comparable to that of CF, and can be used to predict user preferences on new movies. We also investigate the value of movie metadata compared to movie ratings in regards of predictive power. We compare our solely CBF approach with a simple baseline rating-based predictor. We show that even 10 ratings of a new movie are more valuable than its metadata for predicting user ratings.

KW - Collaborative filtering

KW - Content-based filtering

KW - Matrix factorization

KW - Netflix prize

KW - NSVD1

KW - RMSE

UR - http://www.scopus.com/inward/record.url?scp=72249096675&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=72249096675&partnerID=8YFLogxK

U2 - 10.1145/1639714.1639731

DO - 10.1145/1639714.1639731

M3 - Conference contribution

AN - SCOPUS:72249096675

SN - 9781605584355

SP - 93

EP - 100

BT - RecSys'09 - Proceedings of the 3rd ACM Conference on Recommender Systems

ER -