Link to the University of Pittsburgh Homepage
Link to the University Library System Homepage Link to the Contact Us Form

Canonical Correlation Analysis in Cross-Domain Recommendation

Sahebi, Shaghayegh (2016) Canonical Correlation Analysis in Cross-Domain Recommendation. Doctoral Dissertation, University of Pittsburgh.

[img]
Preview
PDF
Download (15MB) | Preview

Abstract

Cross-domain recommendation has recently emerged as a hot topic in the field of recommender systems. The idea is to use rating information accumulated in one domain (known as a source or auxiliary domain) to improve the quality of recommendations in another domain (known as a target domain). One of the important problems in cross-domain recommendation is the selection of source domains appropriate for a target domain. Previous works mostly assume that the best domain pairs can be decided based on similarity of their nature (such as books and movies), or simulate domain pairs by splitting the same dataset into multiple domains.
We argue that the success of cross-domain recommendations depends on domain characteristics and shared (latent) information among domains; therefore posing new questions: What makes a good auxiliary domain? How should we choose the best auxiliary domain for a specific target domain?
In this dissertation we examine the success and failure of cross-domain collaborative filtering across three different datasets with various characteristics of domains. Our goals are to explore the added value of cross-domain recommendations in comparison with traditional within-domain recommendations, and to achieve some progress in uncovering the main mystery of cross-domain recommendation: how can we determine whether a pair of domains is a good candidate for applying cross-domain recommendation techniques? For the former goal, we propose a cross-domain collaborative filtering approach based on canonical correlation analysis. In order to address the latter goal, we investigate a canonical correlation approach as a possible predictor of successful domain pairs and examine a range of features of a single domain and domain pairs in order to see how they could be used to improve predictions. Eventually, we propose a domain-pair classifier that can distinguish between the beneficial and non-beneficial domain pairs before performing the recommendations.


Share

Citation/Export:
Social Networking:
Share |

Details

Item Type: University of Pittsburgh ETD
Status: Published
Creators/Authors:
CreatorsEmailPitt UsernameORCID
Sahebi, ShaghayeghSHS106SHS1060000-0002-8933-3279
ETD Committee:
TitleMemberEmail AddressPitt UsernameORCID
Committee ChairBrusilovsky, Peterpeterb@pitt.ediPETERB0000-0002-1902-1464
Committee MemberCooper, Gregory M.gfc@pitt.eduGMC8
Committee MemberLin, Yu-Ruyurulin@pitt.eduYURULIN
Committee MemberTuzhilin, Alexanderatuzhili@stern.nyu.edu
Date: 3 October 2016
Date Type: Publication
Defense Date: 21 July 2016
Approval Date: 3 October 2016
Submission Date: 10 August 2016
Access Restriction: No restriction; Release the ETD for access worldwide immediately.
Number of Pages: 315
Institution: University of Pittsburgh
Schools and Programs: Dietrich School of Arts and Sciences > Intelligent Systems
Degree: PhD - Doctor of Philosophy
Thesis Type: Doctoral Dissertation
Refereed: Yes
Uncontrolled Keywords: Recommender Systems Cross Domain Collaborative Filtering Domain Analysis Algorithm Canonical Correlation Analysis
Date Deposited: 03 Oct 2016 14:54
Last Modified: 15 Nov 2016 14:35
URI: http://d-scholarship.pitt.edu/id/eprint/29220

Metrics

Monthly Views for the past 3 years

Plum Analytics


Actions (login required)

View Item View Item