Using attributes to improve prediction quality in collaborative filtering

Taek Hun Kim, Sung-Bong Yang

Research output: Contribution to journalArticle

7 Citations (Scopus)

Abstract

To save customers' time and efforts in searching the goods in the Internet, a customized recommender system is required. It is very important for a recommender system to predict accurately by analyzing customer's preferences. A recommender system utilizes in general an information filtering technique called collaborative filtering, which is based on the ratings of other customers who have similar preferences. Because a recommender system using collaborative filtering predicts customer's preferences based only on the items without useful information on the attributes of each item, it may not give high quality recommendation consistently to the customers. In this paper we show that exploiting the attributes of each item improves prediction quality. We analyze the dataset and retrieve the preferences for the attributes because they have not been rated by customers explicitly. In the experiment the MovieLens dataset of the GroupLens Research Center has been used. The results on various experiments using several neighbor selection methods which are quite popular techniques for recommender systems show that the recommender systems using the attributes provide better prediction qualities than the systems without using the attribute information. Each of the systems using the attributes has improved the prediction quality more than 9%, compared with its counterpart. And the clustering-based recommender systems using the attributes can solve the very large-scale dataset problem without deteriorating prediction quality.

Original languageEnglish
Pages (from-to)1-10
Number of pages10
JournalLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume3182
Publication statusPublished - 2004 Dec 1

Fingerprint

Collaborative filtering
Collaborative Filtering
Recommender Systems
Recommender systems
Attribute
Customers
Prediction
Information filtering
Information Filtering
Predict
Experiment
Recommendations
Experiments
Clustering
Internet

All Science Journal Classification (ASJC) codes

  • Theoretical Computer Science
  • Computer Science(all)

Cite this

@article{c7811bc8e5424adb991d2ce1722378c9,
title = "Using attributes to improve prediction quality in collaborative filtering",
abstract = "To save customers' time and efforts in searching the goods in the Internet, a customized recommender system is required. It is very important for a recommender system to predict accurately by analyzing customer's preferences. A recommender system utilizes in general an information filtering technique called collaborative filtering, which is based on the ratings of other customers who have similar preferences. Because a recommender system using collaborative filtering predicts customer's preferences based only on the items without useful information on the attributes of each item, it may not give high quality recommendation consistently to the customers. In this paper we show that exploiting the attributes of each item improves prediction quality. We analyze the dataset and retrieve the preferences for the attributes because they have not been rated by customers explicitly. In the experiment the MovieLens dataset of the GroupLens Research Center has been used. The results on various experiments using several neighbor selection methods which are quite popular techniques for recommender systems show that the recommender systems using the attributes provide better prediction qualities than the systems without using the attribute information. Each of the systems using the attributes has improved the prediction quality more than 9{\%}, compared with its counterpart. And the clustering-based recommender systems using the attributes can solve the very large-scale dataset problem without deteriorating prediction quality.",
author = "Kim, {Taek Hun} and Sung-Bong Yang",
year = "2004",
month = "12",
day = "1",
language = "English",
volume = "3182",
pages = "1--10",
journal = "Lecture Notes in Computer Science",
issn = "0302-9743",
publisher = "Springer Verlag",

}

TY - JOUR

T1 - Using attributes to improve prediction quality in collaborative filtering

AU - Kim, Taek Hun

AU - Yang, Sung-Bong

PY - 2004/12/1

Y1 - 2004/12/1

N2 - To save customers' time and efforts in searching the goods in the Internet, a customized recommender system is required. It is very important for a recommender system to predict accurately by analyzing customer's preferences. A recommender system utilizes in general an information filtering technique called collaborative filtering, which is based on the ratings of other customers who have similar preferences. Because a recommender system using collaborative filtering predicts customer's preferences based only on the items without useful information on the attributes of each item, it may not give high quality recommendation consistently to the customers. In this paper we show that exploiting the attributes of each item improves prediction quality. We analyze the dataset and retrieve the preferences for the attributes because they have not been rated by customers explicitly. In the experiment the MovieLens dataset of the GroupLens Research Center has been used. The results on various experiments using several neighbor selection methods which are quite popular techniques for recommender systems show that the recommender systems using the attributes provide better prediction qualities than the systems without using the attribute information. Each of the systems using the attributes has improved the prediction quality more than 9%, compared with its counterpart. And the clustering-based recommender systems using the attributes can solve the very large-scale dataset problem without deteriorating prediction quality.

AB - To save customers' time and efforts in searching the goods in the Internet, a customized recommender system is required. It is very important for a recommender system to predict accurately by analyzing customer's preferences. A recommender system utilizes in general an information filtering technique called collaborative filtering, which is based on the ratings of other customers who have similar preferences. Because a recommender system using collaborative filtering predicts customer's preferences based only on the items without useful information on the attributes of each item, it may not give high quality recommendation consistently to the customers. In this paper we show that exploiting the attributes of each item improves prediction quality. We analyze the dataset and retrieve the preferences for the attributes because they have not been rated by customers explicitly. In the experiment the MovieLens dataset of the GroupLens Research Center has been used. The results on various experiments using several neighbor selection methods which are quite popular techniques for recommender systems show that the recommender systems using the attributes provide better prediction qualities than the systems without using the attribute information. Each of the systems using the attributes has improved the prediction quality more than 9%, compared with its counterpart. And the clustering-based recommender systems using the attributes can solve the very large-scale dataset problem without deteriorating prediction quality.

UR - http://www.scopus.com/inward/record.url?scp=33645601373&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=33645601373&partnerID=8YFLogxK

M3 - Article

AN - SCOPUS:33645601373

VL - 3182

SP - 1

EP - 10

JO - Lecture Notes in Computer Science

JF - Lecture Notes in Computer Science

SN - 0302-9743

ER -