Collaborative matrix factorization mechanism for group recommendation in big data-based library systems
| Published date | 17 September 2018 |
| Date | 17 September 2018 |
| Pages | 458-481 |
| DOI | https://doi.org/10.1108/LHT-06-2017-0121 |
| Author | Yezheng Liu,Lu Yang,Jianshan Sun,Yuanchun Jiang,Jinkun Wang |
Collaborative matrix factorization
mechanism for group
recommendation in big data-based
library systems
Yezheng Liu, Lu Yang, Jianshan Sun, Yuanchun Jiang and
Jinkun Wang
School of Management, Hefei University of Technology, Hefei, China
Abstract
Purpose –Academic groups are designed specifically for researchers. A group recommendation procedure is
essential to support scholars’research-based social activities. However, group recommendation methods are
rarely applied in online libraries and they often suffer from scalability problem in big data context.
The purpose of this paper is to facilitate academic group activities in big data-based library systems by
recommending satisfying articles for academic groups.
Design/methodology/approach –The authors propose a collaborative matrix factorization (CoMF)
mechanism and implement paralleled CoMF under Hadoop framework. Its rationale is collaboratively
decomposing researcher-article interaction matrix and group-article interaction matrix. Furthermore, three
extended models of CoMF are proposed.
Findings –Empirical studies on CiteULike data set demonstrate that CoMF and three variants outperform
baseline algorithms in terms of accuracy and robustness. The scalability evaluation of paralleled CoMF
shows its potential value in scholarly big data environment.
Research limitations/implications –The proposed methods fill the gap of group-article recommendation
in online libraries domain. The proposed methods have enriched the group recommendation methods by
considering the interaction effects between groups and members. The proposed methods are the first attempt
to implement group recommendation methods in big data contexts.
Practical implications –The proposed methods can improve group activity effectiveness and information
shareability in academic groups, which are beneficial to membership retention and enhance the service
quality of online library systems. Furthermore, the proposed methods are applicable to big data contexts and
make library system services more efficient.
Social implications –The proposed methods have potential value to improve scientific collaboration and
research innovation.
Originality/value –The proposed CoMF method is a novel group recommendation method based on the
collaboratively decomposition of researcher-article matrix and group-article matrix. The process indirectly
reflects the interaction between groups and members, which accords with actual library environments and
provides an interpretable recommendation result.
Keywords Big data analytics, Collaborative matrix factorization, Group recommendation,
Online library system, Personalized services, Scientific article recommendation
Paper type Research paper
1. Introduction
With the rapid popul arity of social network and the increasing development of
information technology, more and more researchers join academic groups in online
library systems. Different fromthenon-academiccommunities,academicgroupsare
designed specifically for scholars. They refer to the virtual communities in which
a group of researchers share academic resources, exchange ideas, follow each
Library Hi Tech
Vol. 36 No. 3, 2018
pp. 458-481
© Emerald PublishingLimited
0737-8831
DOI 10.1108/LHT-06-2017-0121
Received 29 June 2017
Revised 13 December 2017
Accepted 13 December 2017
The current issue and full text archive of this journal is available on Emerald Insight at:
www.emeraldinsight.com/0737-8831.htm
This work is supported by the Major Program of the National Natural Science Foundation of China
(71490725), the Foundation for InnovativeResearch Groups of the National NaturalScience Foundation
of China (71521001), the National Natural Science Foundation of China(71501057, 71722010, 91546114,
71371062) and the National Key Technology Support Program (2015BAH26F00).
458
LHT
36,3
other’s research, keep up with the latest research developments and build up professional
networks (Krause, 2012).
From scholars’perspective, academic groups in online library systems play an
especially vital role in connecting them with each other. In these groups, researchers
collaborative with their colleagues by sharing academic articles and conducting research
discussions (Wei et al., 2015). From service providers’perspective, the library system is
defaulting on its obligation to offer personalized services for individuals or groups
(Renda and Straccia, 2005). An essential factor to the success of group services is group
recommendation procedure, which can improve group activity effectiveness and
shareability of informationinacademicgroups(Kimet al., 2010). Therefore, the
academic group is not only an important academic social network for researchers but also
a basic part of library systems. To satisfy both researchers and online library systems,
it is necessary and worthwhile to explore article recommendation for academic groups in
online library systems.
Numerous group recommenders have been proposed, and they mostly focus on domains
such as TV, movie, music and holidays (Kompan and Bielikova, 2014) rather than online
libraries. Combining individuals’preferences into a group preference (Mccarthy and
Anagnost, 1998; Yu et al., 2006) and aggregating individual recommendations into group
recommendations (Baltrunas et al., 2010; O’Connor et al., 2001) are two typical methods of
group recommendation. However, existing group recommendation methods may not be
applicable to online library systems. First, members in academic groups are often
heterogeneous due to different research-based motivations which impel them to join groups
(Wei et al., 2015). For instance, there is a novice and an expert in a recommender system (RS)
academic group. The former needs classical papers of RSs to learn the foundation, and the
latter requires the newly published papers to get current research trends. Thus,
above-mentioned two group recommendation strategies often result in satisfying the
majority (Kim et al., 2010). Second, existing methods ignore the interactions between group
and members. Third, there are seldom ratings in article recommendation domain compared
with fields like movies and products. Furthermore, they suffer from efficiency and
scalability problems (Meng et al., 2014).
To tackle these issues, we propose a collaborative matrix factorization (CoMF), which
models the preferences of groups and individuals simultaneously by factorizing
researcher-article interaction matrix and academic group- article interac tion matrix.
Researcher-article matrix factorization (MF) is a traditional personalized recommendation
method focusing on individuals’needs, and academic group-article MF is a standard
group recommendation approach focusing on the needs of majority members. Therefore,
we combine both methods to exploit their merits. The collaborative factorization process
of CoMF not only improves the quality of group recommendation in library systems but
also reflects the interaction between group and members. To improve the scalability of
CoMF in big data-based library systems, we implement parallel CoMF on Hadoop
platform based on the MapReduce (Dean and Ghemawat, 2008) parallel processing
paradigm. Furthermore, three extended models of CoMF are proposed by considering the
influence of intergroup similarity and intragroup membership constraint. Finally,
extensive empirical studies demonstrate that CoMF and its three v ariants outper form
baseline algorithms in terms of accuracy and robustness. The scalability evaluation of
paralleled CoMF shows its practice value in big data context.
Our contributions can be outlined as follows: First, CoMF is a novel group
recommendation method by offering appropriate articles to academic groups in online
library systems. It is a vital way to improve the library service, which is advantageous for
enhancing group activity effectiveness and shareability of information in academic groups.
Second, CoMF employs the rationale of collective matrix factorization (CMF) to consider
459
Collaborative
matrix
factorization
To continue reading
Request your trial