Identifying and recommending user-interested attributes with values
| DOI | https://doi.org/10.1108/IMDS-04-2017-0164 |
| Pages | 765-781 |
| Date | 14 May 2018 |
| Published date | 14 May 2018 |
| Author | Yun-Shan Cheng,Ping-Yu Hsu,Yu-Chin Liu |
| Subject Matter | Information & knowledge management,Information systems,Data management systems,Knowledge management,Knowledge sharing,Management science & operations,Supply chain management,Supply chain information systems,Logistics,Quality management/systems |
Identifying and recommending
user-interested attributes
with values
Yun-Shan Cheng and Ping-Yu Hsu
Department of Business Administration, National Central University,
Taoyuan, Taiwan, and
Yu-Chin Liu
Department of Information Management, Shih Hsin University, Taipei, Taiwan
Abstract
Purpose –To retain consumer attention and increase purchasing rates, many e-commerce vendors have
adopted content-based recommender systems. However, apart from text-based documents, there is little
theoretical background guiding element selection, resulting in a limited content analysis problem. Another
inherent problem is overspecialization. The purpose of this paper is to establish a value-based
recommendation methodology for identifying favorable attributes, benefits, and values on the basis of
means-end chain theory. The identified elements and the relationships between them were utilized to
construct a recommender system without incurring either problem.
Design/methodology/approach –This study adopted soft laddering and content analysis to collect
popular elements. The relationships between the elements were established by using a hard laddering online
questionnaire. The elements and the relationships were utilized to build a hierarchical value map (HVM).
A mathematical model was then devised on the basis of the HVM to predict user preferences of attributes.
Findings –The results of a performance comparison showed that the proposed method outperformed the
content-based attribute recommendation method and a hybrid method by 39 and 68 percent, respectively.
Originality/value –Although hybrid methods have been proposed to resol ve the problem of
overspecialization in content-based recommender systems, such methods have incurred “cold start”and
“sparsity”problems. The proposed method can provide recommendations without causing these problems
while outperforming the content-based and hybrid approaches.
Keywords Values, Recommender system, Attribute selection, Means-end chain theory
Paper type Research paper
1. Introduction
To retain consumer attention and increase purchasing rates, many business-to-consumer
vendors implement recommender systems that analyze product attributes and customer
characteristics to predict which products might interest customers (Bobadilla et al., 2013;
Bojnordi and Moradi, 2012; Guo et al., 2017; He et al., 2016; Javari and Jalili, 2014; Mooney
and Roy, 2000; Shih et al., 2002).
Given their simplicity, content-based recommender systems have been in use for a long
time and have been studied by numerous researchers (e.g. Bobadilla et al., 2013; Du et al.,
2011; Protasiewicz et al., 2016). Recommending products using a content-based approach
requires that product attributes be correctly extracted (Pazzani and Billsus, 2007). However,
except for limited types of products, such as books and articles, the extraction process is an
art that relies heavily on experts. This problem is commonly known as the “limited content
analysis problem”(Pazzani, 1999; Pazzani and Billsus, 2007).
An even more serious question is whether physical attributes alone can describe all the
product features that attract customers. For example, on Facebook, many users click or tap
“like”as they view their feeds. Does this mean that the users of social networking sites enjoy
the like function for its own sake or that they wish to express support for friends by
pressing “like”? When social networking sites are recommended to these users, if the answer
is the former, then the like functions of other social networking sites should be
Industrial Management & Data
Systems
Vol. 118 No. 4, 2018
pp. 765-781
© Emerald PublishingLimited
0263-5577
DOI 10.1108/IMDS-04-2017-0164
Received 27 April 2017
Revised 25 July 2017
13 September 2017
Accepted 28 September 2017
The current issue and full text archive of this journal is available on Emerald Insight at:
www.emeraldinsight.com/0263-5577.htm
765
User-interested
attributes with
values
recommended to these users. However, if the answer is the latter, then functions that express
concern for friends should be recommended to these users. Because of this problem, another
inherent issue with content-based approaches is “overspecialization.”The method tends to
recommend very similar items to users and to ignore other items (Adomavicius and
Tuzhilin, 2005; Balabanovićand Shoham, 1997).
Domain ontologies have been proposed to mitigate the limited content analysis problem
(Lops et al., 2011; Middleton et al., 2009), but establishing ontologies requires extensive
expert involvement. Although hybrid approaches that combine content-based and
collaborative filtering methods to mitigate the overspecialization problem have been
proposed, they still suffer from the “cold start”and “sparsity”problems (Blanco-Fernández
et al., 2011; Bobadilla et al., 2013).
The effectiveness of physical attributes has not been formally discussed in the literature
on recommender systems. Studies of the theory of rational action and the technology
acceptance model have consistently confirmed that the intention to use products or services
is driven by the benefits enjoyed by adopting them (Davis, 1989; Kim et al., 2010; Browning
et al., 1999). Means-end chain (MEC) theory posits that when purchasing (or using) a product
(or service), customers (or users) seek benefits and values rather than product attributes
alone. This theory further classifies consumer motivations for purchasing or choosing a
product into benefit and value levels, where values are defined as highly abstract benefits
that summarize desired end-states of being (Gutman, 1982). Benefits are the motivations
that prompt consumers to seek products with the deciding attributes (Gutman, 1982; Haley,
1968; Jung and Kang, 2010; Leisen, 2001; Lin and Lin, 2011; Myers, 1976). In other words,
values are a highly abstract level of motivation that play a dominant role in guiding people
to consider the benefits they desire to derive from the attributes of a product or service
(Goldenberg et al., 2000; Guenzi and Troilo, 2006; Gutman, 1982, 1997; Henneberg et al., 2009;
Klenosky, 2002; Mulvey et al., 1994; Olson and Reynolds, 1983; Reppel et al., 2006; Voss et al.,
2007; Walker and Olson, 1991).
In summary, to design a recommendation methodology that solves limited content
analysis, avoids overspecialization and is not plagued by cold start and sparsity problems,
this paper presents a systematic approach called value-based recommendation (VBR).
The main research goals of this study were as follows:
(1) devise a procedure for systematically extracting attributes, benefits, and values
from products on the basis of MEC theory;
(2) develop an attribute recommendation method on the basis of user values; and
(3) verify the effectiveness of the proposed approach and method by recommending
social networking site attributes on the basis of user values.
To test the effectiveness of the proposed method, user preferences for two major social
networking sites (Facebook and Google+) were studied. The experimental results showed
that the deciding attributes attracting users to either site can be determined using the
proposed procedure and that the attributes recommended using VBR were preferable to
those recommended with content-based and hybrid methods.
The remainder of this manuscript is organized as follows. Section 2 reviews the related
literature, MEC theory, and the laddering approaches used to classify motivations into
various abstract levels. This section also reports on the relationships among attributes,
benefits, and values from the perspectives of the interviewees and discusses the
applications of and issues caused by content-based and hybrid recommender systems.
Section 3 describes the research methodology, including the selection of participants for
soft ladder interviewing who were familiar with both Facebook and Google+,todetermine
the element process and reliability. Section 4 introduces two experiments, as well as
766
IMDS
118,4
To continue reading
Request your trial