4G Mobile Phone Consumer Preference Predictions

by Using the Rough Set Theory and Flow Graphs

Chi-Yo Huang1, Ya-Lan Yang1, Gwo-Hshiung Tzeng2, Shih-Tsung Cheng1, Hong-Yuh Lee3
1
Department of Industrial Education, National Taiwan Normal University, Taipei, Taiwan
2
Department of Business and Entrepreneurial Administration, Kainan University, Taoyuan County, Taiwan
3
Institute of Management of Technology, National Chao Tung University, Hsin-Chu, Taiwan

Abstract--At the moment, when mobile phone users are of mobile phone and make the mobile phone more difficult to
demanding more handset features as well as broader bandwidth, use at the moment when bringing more applicability to
the fourth generation (4G) wireless telecommunication standard consumers [15]. However, at the moment when mobile phone
is emerging. However, how to define appropriate handset vendors complicate the mobile phone designs, the
features toward various market segmentations to fulfill
manufacturing cost is increasing. Meanwhile, the over
customers’ needs and minimize the manufacturing cost has
become one of the most important issues for the 4G handset
complicated features are not necessary what the consumers
manufacturers. Thus, a rule based consumer behavior forecast need. The too high cost may damage mobile phone
mechanism will be very helpful for marketers and designers of manufacturers’ profit margin and thus, their competitiveness.
the handset manufacturers to understand and realize. Moreover, Thus, an analysis of consumer behaviors and predictions
precise prediction rules for consumer behavior being derived by of consumer preferences are essential for mobile phone
the forecast mechanism can be very useful for marketers and manufacturers. Mobile phone vendors should be able to
designers to define the features of the next generation handsets. predict consumers’ preference and design appropriate
Therefore, this research intends to define a Rough Set Theory features to fulfill consumers’ need. Therefore, this research
(RST) based forecast mechanism for the 4G handset feature aims to develop a consumer preference prediction framework.
definitions. Possible handset features will first be summarized To fulfill this purpose, this research intends to develop a
by literature reviews. After that, rules of consumers’ preferences
toward the 4G handsets will be summarized by the RST. To
Rough Set Theory (RST) [17] based 4G mobile phone feature
analyze the data and uncover the knowledge inside the rules definition framework for serving as a ground for mobile
further, the flow graph will further be introduced for analyzing phone manufactures’ 4G mobile phone specification
the information flow distribution. An empirical study on definitions. Moreover, precise “if…, then…” prediction rules
Taiwanese mobile phone users will be leveraged for verifying the for consumer behaviors being derived by the RST forecast
feasibility and demonstrate the usability of the proposed mechanism can be very useful for marketers and designers to
forecast mechanism. Meanwhile, the proposed consumer define the features of the next generation mobile phones. To
behavior forecast mechanism can be leveraged on defining analyze the data and uncover the knowledge inside the rules,
features of other high technology products/services. the flow graph will further be introduced for analyzing the
information flow distribution.
I. INTRODUCTION In this paper, the questionnaires including demographic
variables and consumers’ preferences for certain features
In a relatively short time, the mobile phone has risen to toward the 4G mobile phones were developed. The
become the most common information technology (IT) questionnaires were distributed to users in the Northern part
device in everyday life. A recent United Nations (U.N.) of Taiwan. Based on the survey results, decision rules for
report states that 60% of the world's population has a mobile consumer behavior analysis can be derived by using the
phone subscription [9]. However, the ever-increasing growth RST. To analyze the data and uncover the knowledge inside
of user demand, the limitations of the third generation (3G) of the rules further, the flow graph will further be introduced for
wireless mobile communication systems and the emergence analyzing the information flow distribution. Thus, the mobile
of new mobile broadband technologies on the market have phone manufactures can refer to these rules and flow graphs
brought researchers and industries to a thorough reflection on to develop appropriate marketing, R&D and manufacturing
the fourth generation (4G) [6]. The evolving 4G mobile strategies.
communication systems, which are expected to be launched This research is organized as follows. The related
between 2010 and 2015 [6], are expected to work out literature regarding to consumer behavior, high technology
remaining problems of the 3G systems and to provide a wide market, and the RST applications will be reviewed in Chapter
variety of new services, from high quality voice to 2. The RST based rule derivation method as well as the flow
high-definition video to high-data-rate wireless channels [8]. graph will be introduced in Chapter 3. In Chapter 4, An
According to Norman [15], as mobile phones become empirical study based on the survey data of the 4G mobile
more delicate, more features will be embedded at the same phone consumer behaviors will be introduced to verify the
time. This situation, creeping features, is a tendency to add to feasibility of the proposed framework. Managerial
the number of features that a device can do. Moreover, the implications as well as discussion will be presented in
number of features is usually expanded beyond all reasons Chapter 5. Finally, the whole article will be concluded in
albeit these newly embedded features complicate the design Chapter 6.

978-1-890843-21-0/10/$26.00 ©2010 IEEE

Authorized licensed use limited to: Korea Advanced Inst of Science & Tech - KAIST. Downloaded on March 14,2024 at 05:36:56 UTC from IEEE Xplore. Restrictions apply.
 II. LITERATURE REVIEW According to Barak and Gould [2], consumers of younger
generation are more inclined towards fashionable/stylish
This section shows the basic concepts about the consumer products than older ones. Young consumers are usually more
behavior, consumer behavior in the high-tech market, and the interested in trying out new products and keep getting
RST applications. information about the new ones in contrast to older people.
Their inclinations to get more information make them
A. Consumer Behavior self-confident and are more likely to be opinion leaders. But
Many experts have already proved that the social and older consumers can not be ignored as well. The studies have
psychological factors will affect customers’ behavior [1]. The also showed that older generation consumers are wealthy,
conception of consumer behavior is important for its innovative and they tend to become a part of typical
implication that consumers are different. The idea that only consumption system [3].
some consumers who are interested in certain products are Older consumers can also be a big market for luxury
worth targeting is also essential because it has implied items. But they prioritize other features of the products such
marketing efficiency and effectiveness. Moreover, resource as comfort and convenience higher. It is important to notice
should be spent only on those whom the manager has that older people enjoy their life and are ready to spend their
determined to be potential consumers. Thus, less money is money on later generation people if they feel that the
wasted on consumers who have a low probability of buying a products are relevant [3]. However the younger generation
certain product. It is especially important in an era when people are more acquainted, efficient, adept, ambitious and
competition is being tightened. restless. They always try to look for pleasure and enthusiasm
Understanding the consumer behavior is a vital step in products of the market and different services offered [26].
during making the strategic planning process for the mobile Young buyers’ interest in trying new products, gathering new
phone manufactures. The character of consumer behavior is information makes them more audacious in taking decisions.
about who would buy it, why they would buy it, where they However older consumers are choosy and peculiar about
would buy it, how often they would buy it, and how they innovation and accept few of them which provide certain
would use it [24]. It merges components from, sociology, benefits [14]. Hence, consumer behavior and product
anthropology and economics, etc. The manufactures can selection is highly dependent on age and life cycle [11].
profit from an understanding of consumer behavior, thus they The task which a marketing manager is facing is to
can make better prediction for what consumers would want to identify variables that describe the consumers in terms of
buy and how to do the best to fit the consumers’ need. their inherent characteristics to link those variables to
The study of consumer behaviors helps mobile phone consumers and certain products. Moreover, the marketer of
manufactures develop their marketing strategies by mobile phone manufactures can use most the major factors,
understanding issues including: (1) how consumers think, feel, such as age, gender, income level, and the frequency of
and choose from different products; (2) how the consumers switching the mobile phone, to outline the decision rules of
are influenced by the environment and their background (e.g., what distinct consumer segmentations will buy the 4G mobile
media, family); (3) the behavior of consumers while making phones. Understanding of these influences may help mobile
buying decisions; (4) the decisions influenced by the phone manufactures identify the groups of consumers who
limitation of consumer information abilities; (5) how tend to feel, think, or act similarly and separate them into
consumers’ decision strategies and motivation differ between different unique market segmentations. Aspects of consumer
products that are different in level of interest or importance; behavior can be tailored to meet those unique needs, values,
and (6) how company can adapt and improve their marketing and goals of these distinct consumer segmentations.
strategies to more effectively achieve the consumers’ need
[21]. B. Consumer Behavior in High technology Market
In 2002 Sabnavis [23] classified consumers into three Comparing to the common products, mobile phone is in a
different categories for three generations. The consumers of special market, a part of the high technology market, which
the first generation (1960-70s) were traditional in nature, demonstrates following characteristics. (1) The marketing
balanced, inward looking and had reduced choices. For them, strategy of high-tech product is very different from other
the needs of the family were the top priority and later came common products because it exists insecurity in consumers
their personal needs. They abstained themselves from taking and the adoption process of high-technology product is
risks. However, the consumers of the second generation distinct. (2) Predicting consumer’s satisfaction is tough for
(1980-90s) took risks and chances more than their high technology products because these products are more
forerunners. They encountered multiple choices and had an complicated than common products. therefore, to achieve
impulse of being better off than their parents. In addition to customers’ satisfaction, the high technology manufactures
that, they had very little fears and worries economically. The need to inform and support consumers with broad-ranging
consumers of the third generation (new millennium) enjoy levels of expertise and information. (3) As a result of network
life to their fullest. They have more control on themselves externalities, high technology are managed by a small part of
and personal style and pleasure is very important for them. vital mass, often lead to a “winner-takes-all” outcome [16]. (4)

Authorized licensed use limited to: Korea Advanced Inst of Science & Tech - KAIST. Downloaded on March 14,2024 at 05:36:56 UTC from IEEE Xplore. Restrictions apply.
 The life cycles of high tech market is very short. Moreover, III. METHODS
there is considerably more pressure on marketing managers to
make profits as quickly as possible. (5) The market is In Chapter 3, we will introduce the RST and the Flow
turbulent. The high-tech manufactures will be found and Graph for analyzing customers’ preference toward the 4G
disappear weekly with profit going up or down rapidly. mobile phone by deriving the “if…then” rules as well as
Further, as summarized by Winer [30], there are three establishing the flow graphs for demonstrating the rules
issues which high tech manufactures will probably face. (1) further. In Section 3.1, the RST is described. In Section 3.2,
Consumer issues: who are the potential consumers for the the Flow Graph is introduced.
high technology products? How much do they recognize
about the service features and the benefits? How much are A. Rough Set Theory
they are willing to spend? (2) Strategic issues: assuming that The Rough Set Theory being proposed by Pawlak [17] in
the high tech manufactures decides to launch a new service, 1982 is a new mathematical approach to handling vagueness
what strategy should be used? (3) Competitive issues: which in data. The Rough Set philosophy is founded on the
manufactures and substitute technologies will the high tech assumption that with every object of the universe of discourse,
manufactures compete against? some information (data, knowledge) can be associated. For
Regarding new high technology, some consumers are risk example, if objects are patients suffering from a certain
takers and others are risk averse. Generally, the rate of disease, symptoms of the disease form information about
adoption of new high technology is highly reliant on how patients. Objects characterized by the same information are
many consumers in the early adopter segmentation are indiscernible (similar) in view of the available information
classified by the “Diffusion of Innovation Theory”; if that about them. The indiscernibility relation generated in this
segmentation grows, the technology is probably to be adopted way is the mathematical basis of the RST [19].
by a large fraction of the potential consumer segmentation. The RST seems to be of fundamental importance to A.I.
Thus, the high tech manufactures should take advantage of and cognitive sciences, especially in the areas of machine
the early adopter segmentation, consumers are willing to learning, knowledge acquisition, decision analysis,
adopt the new high technology, to enter the mass market, and knowledge discovery from databases, expert systems,
the new technology will be expected to be successful. inductive reasoning and pattern recognition. The RST has
One of the interesting questions for marketer of the high been successfully applied in many real-life problems in
tech manufactures is how to identify which segmentations medicine, pharmacology, engineering, banking, finances,
should the consumers be classified in? Moreover, the market analysis, environment management and others [19].
consumer segmentation which has gained the most attention This research uses the definitions of the RST introduced
is the earlier adopter segmentation. As summarized by Perner, by Walczak and Massart [29] as indicated below.
Alpert and Kamins [21], some traits have been associated
with that segmentation: (1) higher income, (2) higher Step1: Information system:
education, (3) age, (4) greater social mobility, (5) Formally, an information system denoted IS is defined
venturesomeness, (6) greater social participation, and (7) high by IS = (U , A ) , where U consists of finite objects and is
opinion leadership [21]. High technology manufactures, like named a universe and A is a finite set of attributes {a1 , a2 ,...,an } .
4G mobile phone manufacture, can follow these traits to Each attribute a∈ A defines an information
discover the earlier adopter consumers to raise the
function f a : U → Va , where Va is the set of values of a, called
successfulness of a new technology.
the domain of attribute a.
C. The RST Applications
The advantages of RST are that it does not need any Step2: Indiscernibility relation:
hypothesis and external information. Furthermore, it can deal For every set of attributes B ⊂ A , an indiscernibility
with vagueness and uncertainty of information [29]. The RST relation Ind(B) is defined in the following way: two objects,
offers an alternative toolset for business analysis. The xi and x , are indiscernible by the set of attributes B in A, if
j

practice of using rough sets for decision rule generation has b( xi ) = b( x ) for every b ⊂ B . The equivalence class of
j

increased in popularity recently [31]. Ind(B) is called elementary set in B because it represents the
The RST has also been used for marketing decision smallest discernible groups of objects.
support systems in Swedish retail business [12] so as to
capture the relationships between sales levels with marketing Step 3: Lower and upper approximations:
factors, such as advertising and promotion. Other applications Let X be a subset of elements in universe U, that
include tourism shopping [13], business failure prediction [5], is, X ⊂ U . Let us consider a subset in Va , B ⊆ Va . The low
etc.
approximation of B, denoted as BX , can be defined by the
union of all elementary sets X i contained in X as follows:

Authorized licensed use limited to: Korea Advanced Inst of Science & Tech - KAIST. Downloaded on March 14,2024 at 05:36:56 UTC from IEEE Xplore. Restrictions apply.
 {
BX = xi ∈ U ⎣⎡ xi ⎦⎤ Ind(B) ⊂ Χ } (1) B ( F ) = {B( X 1 ), B( X 2 ),..., B( X n )}

where X i is an elementary set contained in X, i = 1,..., n . The accuracy of the classification is:
β B F = ∪cardB( X i ) / ∪cardB ( X i ) (6)
The upper approximation of B, denotes as BX, can be
denoted by a non-empty intersection of all elementary sets The quality of the classification is:
X i contained in X as follows: η B F = ∪cardB( X i ) / cardU (7)

{
BX = xi ∈ U ⎡⎣ xi ⎤⎦ Ind(B) ∩ Χ ≠ ∅ } (2)
Step 7: Decision table:
The boundary of X in U is defined in the following: A knowledge representation system containing the set of
BNX=BX-BX (3) attributes A and the set of decision attributes D is called a
decision table; decision tables are also useful for
Step 4: Independence of attributes: classification. There are four main steps of decision table.
In order to check, whether the set of attributes is 1. Construction of elementary sets in D-space,
independent or not, one checks for every attribute whether its 2. Calculation of upper and lower approximations of the
removal increases the number of elementary sets in the IS or elementary sets in D,
not. 3. Finding D-core and D-reducts of A attributes,
If Ind ( A) =Ind ( A − ai ) , then the attribute ai is called 4. Finding D-core and D-reducts of A attribute values.
superfluous. Otherwise, the attribute ai is indispensable in
Step 8: Decision rules:
A. Decision rules can also be regarded as a set of decision
(classification) rules of the form: ak = d j . Where ak means
Step 5: Core and reduct attributes: i i

Core and reduct of attribute sets are two fundamental that attribute ak has value i, d j means the decision
concepts of a rough set. Reduct is a minimal subset and attributes and the symbol ‘ ⇒ ’ denotes propositional
makes the object classification the full set of attributes. The implication. In the decision rule Φ ⇒ Ψ , formulas Φ and
core is common to all reducts. Reduct attributes can remove Ψ are called condition and decision, respectively.
the superfluous attributes and provide the decision maker a Minimization of a set of attributes and values of attributes
simple and useful information. There may be more than one with respect to another set of attributes simply means a
reduct attributes. If the set of attributes is dependent, we are reduction of unnecessary conditions in the decision rules,
interested in finding all possible minimal subsets of attributes which is also known as the generation of decision rules from
which have the same number of elementary sets; it’s called the data.
the reducts. Through the decision rules we can minimize the set of
The reduct attribute set affects the process of attributes, reduct the superfluous attributes and group
decision-making, and the core attribute is the most important elements into different group. In this way we can have many
attribute in decision-making. If the set of attributes is decision rules, each rule has meaningful features. The
indispensable, the set is called the core. stronger rule will cover more objects and the strength of each
RED(B) ⊆ A (4) decision rule can be calculated in order to decide the
COR(C ) = ∩ RED(B) (5) appropriate rules.
To compute reducts and core, the discernibility matrix is
used. The discernibility matrix has the dimension n × n , Step 9: The measures of quality in classification:
where n denotes the number of elementary sets and its The suppS (Φ,Ψ ) is called the support of the rule Φ → Ψ
elements are defined as the set of all attributes which discern in IS and card(U ) is the cardinal set that is the number of
elementary sets ⎣⎡ x ⎦⎤i and ⎣⎡ x ⎦⎤ j . objects contained in the U [14].
σ S (Φ,Ψ ) = suppS (Φ,Ψ ) / card(U ) (8)
Step 6: Classification: is the strength of the decision rule Φ → Ψ in S.
Let F = { X 1 , X 2 ,..., X n } , X i ⊂ U , be a family of subsets of covS (Φ,Ψ ) = suppS (Φ,Ψ ) / card(Ψ S ) (9)
the universe U. If the subsets in F do not overlap, is the coverage factor of the decision rule Φ → Ψ in S.
i.e., X i ∩ X j = ∅ , and the entity of them contains all
elementary sets, i.e., ∪ X i = U for i = 1,..., n . Then, F is B. Flow Graphs
Flow graphs, proposed by Pawlak [19], can be considered
called a classification of U, whereas X i are called classes. as a special kind of databases, where instead of data about
The lower and upper approximations of F in B ⊆ A are individual objects some statistical features of objects are
defined as: presented in terms of information flow distribution. It turns
B ( F ) = {B( X 1 ), B( X 2 ),..., B( X n )} out that such data representation gives a new insight into data
structures and leads to new methods of intelligent data

Authorized licensed use limited to: Korea Advanced Inst of Science & Tech - KAIST. Downloaded on March 14,2024 at 05:36:56 UTC from IEEE Xplore. Restrictions apply.
 analysis. normalized inflow
This research uses the definitions of the flow graphs ϕ + ( x)
ϕ + ( x) = = ∑ σ ( y , x) (15)
introduced by Pawlak [20] as indicated below. ϕ (G ) y∈I ( x )
A flow graph is a directed acyclic finite graph and normalized outflow
G = ( N , β ,ϕ ) , where N is a set of nodes, β ⊆ N × N is a set of
ϕ − ( x)
directed branches, ϕ : β → R + is a flow function, and R+ is the ϕ − ( x) = = ∑ σ ( x, y ) (16)
ϕ (G ) y∈O ( x )
set of non-negative reals. It can be listed basic concepts of
Obviously, for any internal node x, it
flow graphs:
has σ + ( x) = σ − ( x) = σ ( x) , where σ ( x) is a normalized
• If ( x, y ) ∈ β then x is an input of y and y is and output of
through flow of x. Moreover, let
x. ϕ + (G )
• If x ∈ N then I(x) and O(x) denote the sets of all x’s inputs σ + (G ) = = ∑ σ − ( x) (17)
ϕ (G ) x∈I ( G )
and outputs.
ϕ − (G )
• Input and output of a graph G are defined, respectively, as σ − (G ) = = ∑ σ + ( x) (18)
I (G ) = {x ∈ N : I ( x) = ∅} and O(G ) = {x ∈ N : O ( x) = ∅} ϕ (G ) x∈O ( G )
• Inputs and outputs of G are its external nodes; other nodes Obviously, σ + (G ) = σ − (G ) = σ (G ) = 1 .
are internal. A (directed) path from x to y, x ≠ y , in G is a sequence of
• If ( x, y ) ∈ β the ϕ ( x, y ) ⊇ is a through flow from x to y; nodes x1 ,..., xn such that x1 = x , xn = y and ( xi , xi +1 ) ∈ β for every
• It will be assumed in what follows that ϕ ( x, y ) ≠ 0 â for i, 1 ≤ i ≤ n − 1 . A path from x to y is denoted by [x…y] and n-1
is called length of the path.
every ( x, y ) ∈ β .
A flow graph is linear if all paths from node x to node y
• With every node x of a flow graph G it can be associated have the same length, for every pair of nodes x, y.
its inflow
ϕ + ( x) = ∑ ϕ ( y , x) (10) IV. PREDICTING THE PREFERENCE OF CONSUMERS
y ∈I ( x )
TO 4G MOBILE PHONE
and outflow
ϕ − ( x ) = ∑ ϕ ( x, y ) (11) A. The History of Mobile Phone
y∈O ( x )
In this section, the history and evolution of mobile phone
Similarly, it can be defined an inflow and an outflow for technology will be discussed for serving as a background for
the whole flow graph G: the RST based mobile phone consumer behaviors analysis in
ϕ + (G ) = ∑ ϕ − ( x), (12) Section 4.5. All four generations of technology as well as
x∈I ( G )
applications of the 4G phones will be introduced.
ϕ − (G ) = ∑
x∈O ( G )
ϕ + ( x). (13) In 1970s the mobile phone technology started also called
as 1G. The very first systems were very different from
We assume that for any internal node
today’s, instead of digital, analog systems formed the basis of
x, ϕ + ( x) = ϕ − ( x) = ϕ ( x) , where ϕ ( x) is the through flow of node
this technology and the mobile structure of mobile
x. Obviously, ϕ + (G ) = ϕ − (G ) = ϕ (G ) , where ϕ (G ) is the through communication. A lot of basic problems of the consumers
flow of G. were solved by this 1G system, however, during 1980s, a lot
The above formulas can be considered as flow of incompatible analog systems were used in service around
conservation equations [5]. N is a set of nodes, β ⊆ N × N is a the world.
set of directed branches, but instead of ϕ : β → R + it has a The main difference in technology came with the
normalized flow defined by introduction of 2G (second generation) systems designed in
ϕ ( x, y ) the 1980s. These systems were used for voice applications
σ ( x, y ) = (14) just like 1G. However, these were based on digital
ϕ (G )
technology, including digital signal processing techniques.
for any ( x, y ) ∈ β . The function of the 2G systems was to provide circuit-
The value of σ ( x, y ) is called the strength of (x, y). switched data communication services but at a low speed.
Increasing number mobile phone manufactures and a race to
Obviously, 0 ≤ σ ( x, y ) ≤ 1 . The strength of the branch
design and implement digital systems resulted in a lot of
expresses simply the ratio of through flow of the branch to different and incompatible standards such as GSM (global
the total flow. system mobile), mainly in Europe; TDMA (time division
Normalized graphs have interesting properties which are multiple access) (IS-54/IS-136) in the U.S.; PDC (personal
discussed next. In what follows it will be used normalized digital mobileular) in Japan; and CDMA (code division
flow graphs only. multiple access) (IS-95), another U.S. system. Digital
With every node x of a flow graph G we associate its systems function all around the country as well as

Authorized licensed use limited to: Korea Advanced Inst of Science & Tech - KAIST. Downloaded on March 14,2024 at 05:36:56 UTC from IEEE Xplore. Restrictions apply.
 internationally are today's leading systems, although the data wireless networks at very high speeds, as high as 100
rate for users in systems is very limited. Mbps, about 260 times than 3G wireless network.
It was in 1990s that two manufactures tried to work (2) Interoperability: One problem with 3G is not being able to
together to come up with 3G systems which would be free operate across different networks, thus not able to become
from all incompatibilities of previous 2G systems and a truly global technology. What we need is a global
becomes global system. The 3G system would not only have standard which would provide operation among different
better voice quality but also broadband data capabilities even networks across different countries providing a global
up to 2Mbps. Unfortunately, the two manufactures were not mobility and service portability. This would not only help
able to overcome their differences, and therefore this decade consumers but also help provider which would no longer
will see the introduction of two mobile standards for 3G. be bound by single-system vendors of proprietary
Also, China is at the brink of implementing of third 3G equipment.
system. (3) Networking: 3G technology used only WAN concept but
An intermediate step is also being taken into limited on a networking scale. A hybrid technology of
consideration which is between 2G and 3G, also called as the networks that utilizes both wireless LAN (hot spot)
2.5G. It is an upgrade of two major 2G technologies which concept and mobile or base-station WAN design is needed
would provide higher capacity on the 2G RF (radio frequency) today. 4G would be based on different stations
channels and also implement larger throughput for data everywhere would provide global connection anytime
service, up to 384 kbps. Optimization of data channels for anywhere to mobile phone users.
packet data is also an extremely important aspect of 2.5G (4) Bandwidth: A higher bandwidth is needed to access data
which would implement access to the Internet from mobile across networks at higher speeds. The 4G network will
devices, whether telephone, PDA (personal digital assistant), provide speeds of more than 20 mbps, thus offering
or laptop. However, in today’s world the need to higher data high-bandwidth services within the reach of LAN
transfer speed is increasing which is mainly for multimedia "hotspots," installed in different places like offices, homes,
communication and it basically depends on computer coffee shops, and airport lounges. When consumers are
communication in digital format. According to the historical away from these LAN hotspots, they can connect to slow
indication, revolution in the mobile systems technology by 2G networks for voice and rudimentary data coverage.
introduction of new generation mobile phone occurs every (5) Technology: Unlike 3G, 4G will be more of a
decade, therefore, this present appears to be the right time to combination of older technologies and not totally a new
begin the research on a 4G mobile communication system. standard. According to analysts 4G technology is a
seamless combination of existing 2G wireless networks
B. 4G mobile phone with local-area networks (LANs) or Bluetooth.
4G is the latest generation of mobile communication (6) Convergence: Convergence is the coming together of
system that will modify and replace 3G systems in future 5 to markets and services and not just technologies. Different
10 years [10]. 4G is proposed to offer high capacity, high networks that utilize IP in its fullest form with converged
speed, IP based services, low cost per bit, etc [25]. One term voice and data capability is needed, which the 4G will
can describe 4G is MAGIC- “Mobile multimedia, Anytime achieve.
anywhere, Global mobility support, Integrated wireless (7) Cost: The cost of 4G technology would be less because it
solution, and Customized personal service.” [10] would be based on combination of existing technologies
and it would work on existing networks and hence no new
C. Evolution from 3G to 4G tools will be needed by the operators and won't require
It is a very difficult phase for 3G networks because of carriers to purchase costly extra spectrum. Costs will be
huge costs, technical issues and network operators deflating further reduced by the fact that 4G technology would be
expectations based on unrealistic hype all leading to failure of based on an open system IP wireless environment by
3G. Despite the hype encompassing high speed of 3G ushering in an era of real equipment interoperability.
networks is building up, Santhi et al. [25] summarized the (8) Scalability: With increasing number of users, the
reasons for the leap towards 4G as (1) performance, (2) technology should be able to handle more users, hence
Interoperability, (3) Networking, (4) Bandwidth, (5) should be scalable. "Design for Scalability," includes
Technology and (6) Convergence, (7) cost, and (8) Scalability. information that can help you meet changing usage
These reasons are explained further below based on the work demands. The 4G network is best in this scenario because
by Santhi et al. [25]. it is based on all IP core layers which is easily scalable
(1) Performance: Mobile phone industry critics estimate that and hence will be able to meet the challenge of scalability.
mobile phone users will not be able to enjoy very high
speeds thus refraining them from rich multimedia content D. Applications of 4G
across wireless networks with 3G. 4G communications The 4G mobile phone is supposed to provide more
will not only give very high-quality video equal to that of convenient applications, from 3D virtual image system to
high definition television but also enable downloading on wider bandwidth wireless channels, in our daily life. The

Authorized licensed use limited to: Korea Advanced Inst of Science & Tech - KAIST. Downloaded on March 14,2024 at 05:36:56 UTC from IEEE Xplore. Restrictions apply.
 applications of 4G mobile phone were described as follow. (1) questionnaires were distributed to customers in the Northeast
Smart Remote Control: With your whole home devices are districts of Taiwan. 56 (84%) valid questionnaires from total
connected to the mobile network, your mobile phone will 67 were received. Moreover, the four-step analytical
become the ultimate remote controller, and let you control procedure being provided by the ROSE2 (Rough Set Data
every device in your home, from your air-conditions, washing Explorer) being developed by Predki et al. [22] which can
machine, oven, even your light-bulbs, regardless of where derive the decision rules based on the RST. The analytic
you are in the world. (2) Wider Bandwidth and Higher bit procedures are demonstrated below.
Rate: Data sharing among the different networks available
will solve the problem related to the spectrum limitations Step 1: Attributes domain definition and the decision
compared with 3G [6]. Wider bandwidth will solve the table establishment
problems of limitation in 3G when people are on the The information system S, in view of previous analysis is
movement while using the wireless system, and also used to process attributes for extraction, and finally adds the
providing bandwidth in every locations. (3) Virtual 3D image survey results of personal attributes. Hence a total of nine
technology: 4G will provide 3D virtual reality; the ability to attributes including eight condition attributes and one
sense as if you are there at an event while actually you are not. decision attribute were derived. The eight condition attributes
Its applications are found in fields such as national defense, include gender (a1), age (a2), monthly income (a3),
medicine, education, and gaming. (4) Higher Performance accumulated frequency of switching mobile phone (a4), smart
Multimedia: 4G can support streamed HD television – remote control (c1), wider bandwidth and higher bit rate (c2),
download movie just need around 5 minutes - as well as virtual 3D image technology (c3), and higher performance
advanced MMS (multimedia messaging service), and video multimedia (c4). The decision attribute is the preference of
chat - all delivered 4G spirits “anytime, anywhere. customers toward 4G mobile phones. Two decision classes
including decision class 1, (the consumer will purchase 4G
E. An Empirical Study of the 4G Mobile Phone Consumer mobile phones) and decision class 2 (the consumer will not
Behavior purchase a 4G mobile phone)– expressed different tendencies
At first, the questionnaires included the demographic of of customers’ preference toward 4G mobile phones. The
variable and consumers’ preference for certain features attributes domain and the data are summarized in Table 1.
toward the 4G mobile phones were designed. The
TABLE 1. ATTRIBUTE SPECIFICATION FOR THE PERSONAL ANALYSIS
Attribute Name Attribute Values Attribute Value Sets
Condition attributes
Gender (a 1 ) Male; Female {1,2}
Age (a 2) <30; 31~35; 36~45; 46~ {1,2,3,4}
Monthly Income (a 3) <10 thousands; 11~30 thousands; 31~50 thousands; 51thousands~ {1,2,3,4}
Accumulate frequency of switching mobile phone (a 4) 1~2 times; 3~4 times; 4~5 times; 6 times~ {1,2,3,4}
Smart Remote Control (c 1) Very strongly important; Moderate important; Not important {1,2,3}
Wider Bandwidth and Higher bit Rate (c 2) Very strongly important; Moderate important; Not important {1,2,3}
Virtual 3D image technology (c 3) Very strongly important; Moderate important; Not important {1,2,3}
Higher Performance Multi-Media (c 4 ) Very strongly important; Moderate important; Not important {1,2,3}
Decision attributes
Preference of customers for 4G mobile phone (d ) Will buy; will not buy {1,2}

Step 2: Approximation calculation Step 3: The reducts of attributes and the core of attributes
The approximations of decision classes were first derivations
calculated by using the ROSE2. As shown in Table 2, the By employing the discernibility matrix, this research
decision class 1 was well described due to its high accuracy derived all potential reducts in the decision table. Moreover,
of 0.875 while the accuracy of decision class 2 is 0.85. On the only three reducts of attributes and four cores of attributes
whole, the accuracy of the entire classification is 0.8625 were derived as the result. Three reduct sets of attributes
while the quality of the entire classification is 0.9268. While {a2,a3,a4,c2,c3,c4}, {a2,a3,c1,c2,c3,c4}, {a3,a4,c1,c2,c3,c4} and
the accuracy is higher, the classification is less ambiguous. four cores {a3,c2,c3,c4} were derived. It implies that the
While the quality is higher, the classification is better [21]. relationships between attributes, the most important attribute,
or the least important attribute of the data can be determined
TABLE 2. THE LOWER AND UPPER APPROXIMATIONS by the RST. Therefore, attributes a2, a3, a4, c1, c2, c3, and c4
Class Number 1 2 are the most meaningful attributes among the eight attributes
Number of Objects 23 18
Lower Approximation 21 17 while a1 is the superfluous attribute.
Upper Approximation 24 20
Accuracy 0.875 0.85

Authorized licensed use limited to: Korea Advanced Inst of Science & Tech - KAIST. Downloaded on March 14,2024 at 05:36:56 UTC from IEEE Xplore. Restrictions apply.
 Step 4: Developing the decision rules TABLE 4. RULES BEING DERIVED CORRESPONDING TO THE
DECISION VARIABLE 1
In this study, 20 rules were derived by the ROSE2. The
Decision
decision rules and specific observations conform some Rule a1 a2 a3 a4 c1 c2 c3 c4
Variable
Support

decision rules are shown in Table 3. 1 4 4 1 6

TABLE 3. DECISION RULES GENERATED FROM ROSE2 2 2 1 4

Decision
Rule a1 a2 a3 a4 c1 c2 c3 c4 Strength(%)
Variable 3 2 3 1 2
1 4 4 1 14.63
2 2 1 9.76 4 3 1 1 6
3 2 3 1 4.88
4 3 1 1 14.63 5 2 1 1 1 1 2
5 2 1 1 1 1 4.88
6 3 2 1 2 1 2.44 6 3 2 1 2 1 1
7 1 3 1 2.44
8 3 2 1 1 7.31 7 1 3 1 1
9 1 2 1 1 4.88
10 1 2 2 2 2.44
8 3 2 1 1 3
11 2 2 1 2 4.88
12 1 3 2 4.88
9 1 2 1 1 2
13 1 2 2 9.76
14 1 3 2 9.76
20 4 1 1 2
15 3 2 2 4.88
16 2 3 2 4.88
17 2 4 2 2 4.88
18 3 2 2 2 2.44 The decision rules of decision variable 2 is shown in Table
19 1 1 2 2 4.8 5. The rules can be translated into one decision algorithm
20 4 1 1 or 2 7.31
represented by the decision flow graph as shown in Fig. 2
which demonstrate some common characteristics of the
Step 5: Using the flow graphs consumers who will not purchase a 4G mobile phone (d=2):
The rules in Table 3 can be translated into two decision (1) female (a1=2), (2) young age (a2=1, 2), (3) moderate
algorithms being represented by the decision flow graphs income (a3=3), (4) low frequency to switch phones (a4=1), (5)
which are shown in Fig. 1 and Fig. 2. Moreover, the decision ignorance of novel features including on the smart remote
rules of decision variable 1 is as shown in Table 4 that can be control, wider bandwidth and higher bit rate, higher
translated into one decision algorithm being represented by performance multimedia (c1=2, c2=2, c4=2), and the virtual
the decision flow graph as shown in Fig. 1. For simplicity, the 3D image technology (c3=3).
strength being associated with each branch is omitted.
By using the Fig. 1, decision makers can easily understand TABLE 5. RULES BEING DERIVED CORRESPONDING TO THE
which values of condition attributes appear frequently. DECISION VARIABLE 2
Decision makers may uncover the characteristics of the Rule a1 a2 a3 a4 c1 c2 c3 c4
Decision
Support
Variable
consumers willing to buy the 4G mobile phones (d=1) as
follows: (1) female (a1=2), (2) the age between 36 to 45 10 1 2 2 2 1

years (a2=2), (3) high monthly income (a3=4), (4) the 11 2 2 1 2 2

frequency to change mobile phones is above average (a4=2,
4), (5) focus on smart remote control, wider bandwidth and 12 1 3 2 2
higher bit rate, and higher performance multimedia (c1=1,
13 1 2 2 4
c2=1, c4=1), as well as (6) ignorance of the virtual 3D image
technology (c3=3). 14 1 3 2 4

15 3 2 2 2

16 2 3 2 2

17 2 4 2 2 2

18 3 2 2 2 1

19 1 1 2 2 2

20 4 1 2 1

Authorized licensed use limited to: Korea Advanced Inst of Science & Tech - KAIST. Downloaded on March 14,2024 at 05:36:56 UTC from IEEE Xplore. Restrictions apply.
 1

2 a3=1 a4=1 2
2
2 a2=1 2 c4=1
2
1 2 2 4
a3=2 a4=2 2 4
4 6 4 1
29 2 3 2 c1=1 c2=1 c3=3 d=1 6
a1=2 a2=3 c4=2
1 6
9 1 1 2
a4=3 3
a3=3 1
a2=4 c4=3
2 6 1
a3=4 a4=4
6

Fig. 1: Decision Flow Graph for Decision Variable 1

6 2
1
a1=1 2 a2=1 1 a3=1 a4=1 c3=1
4 2
2
c2=2 2 2 3
4 2 1
2 a2=2 2 1 d=2
23 a3=3 a4=2 4 c1=2 c3=2 4
2 c4=2 6
2
2 1 1 4
1
1
a1=2 a2=4 a3=4 a4=3 c3=3
6 2
2

Fig. 2: Decision Flow Graph for Decision Variable 2

V. DISCUSSION well predict consumer behaviors by using the two rules

further.
This section tends to discuss both managerial implications Finally, the strength of the most supported rules is 14.6%
as well as advances in research methods. For the rules being only. Most rules are supported merely by very little objects.
derived, rule 1 to rule 9 can be used to predict the behaviors This phenomenon implies the high uncertainty nature of the
of consumers who will purchase the 4G mobile phones. The high technology marketing environment. Therefore, all of the
rules 10 to 19 can used to predict the behaviors of consumers rules have the low strength. However, the RST can still be
who will not purchase a 4G mobile phone. The exception rule, leveraged to derive rules which cover only subsets of the
number 20, is the approximate rule which overlaps more basic objects or data records available [2]. Given that, the
than one decision class. RST is undoubtedly suitable to derive the preference of
By using the RST, decision makers of mobile phone consumers in the market of high technology products.
manufactures can easily understand behaviors and
characteristics of the consumers willing to purchase 4G VI. CONCLUSIONS
mobile phones. Taking the decision rule 5 (please refer to
Table 3) as an example, a young (a2=1) female (a1=2) The 4G wireless telecommunication standard is emerging
preferring high wireless communication bandwidth (c2=1) at the moment, when mobile phone users are demanding
and high performance multimedia processing (c4=1) will more handset features as well as broader bandwidth. However,
probably buy 4G mobile phones (d=1). how to define appropriate handset features toward various
In this research, the most supported rules are rule 1 and market segmentations to fulfill customers’ needs and
rule 4 (strength = 14.6 %). According to the rule 1,the high minimize the manufacturing cost has become one of the most
income consumers (a3=4), who change handset frequently important issues for the 4G handset manufacturers. In this
(a4=4) will probably purchase the 4G mobile phone (d=1). research, a consumer behavior prediction framework
According to the rule 4, middle-income consumers requiring consisting of the RST and flow graphs is proposed. An
smart remote control features will probably purchase 4G empirical study on the northern Taiwan’s consumer
mobile phones. Thus, 4G Mobile phone manufacturers may preferences toward 4G mobile phones was leveraged for

Authorized licensed use limited to: Korea Advanced Inst of Science & Tech - KAIST. Downloaded on March 14,2024 at 05:36:56 UTC from IEEE Xplore. Restrictions apply.
 verifying the feasibility of the framework. Meanwhile, erived 241-249, 2000.
[14] Nam, J., Hamlin, R., Gam, H. J. Kang, J. H., Kim, J., Kumphai, P.,
20 prediction rules were derived by the analytic framework, Starr, C. and Richards, L. The fashion-conscious behaviours of mature
These rules can serve as the basis of consumer behavior female consumers. International Journal of Consumer Studies, 31 (1),
predictions by 4G mobile phone manufactures in the future. pp. 102-108, 2007.
Further, the proposed framework can also be leveraged on [15] Norman, D.; “The Design of Everyday Things,” New York: Basic
Books, pp. 172, 1988.
predicting consumer behaviors of other high technology [16] Parasuraman, A.; “Technology Readiness Index (TRI): A
products. Multiple-Item Scale to Measure Readiness to Embrace New
Technologies,“ Journal of Service Research, 2 (4), pp. 307-320, May
REFERENCES 2000.
[17] Pawlak, Z.; “Rough sets,” International Journal of Computer and
Information Sciences, 11 (5), pp. 341-356, 1982.
[1] Ajzen, I. and M. Fishbein, Understanding attitudes and predicting
[18] Pawlak, Z.; “Rough sets and intelligent data analysis,” Information
social behavior. Englewood Cliffs, NJ: Prentice-Hall, Inc, 1980.
Sciences, 147 (1-4), pp. 1-12, 2002a.
[2] Barak, B. and Gould, S.. Alternative age measures: a research agenda.
[19] Pawlak, Z.; “Rough sets, decision algorithms and Bayes’theorem,”
Advances in Consumer Research, 12, pp.53-58, 1985.
[3] Carrigan, M. and Szimigin, I. . In pursuit of youth: what’s wrong with European Journal of Operational Research, 136, pp. 181-189, 2002b.
the older market? Marketing Intelligence and Planning, 17(5), pp. [20] Pawlak, Z.; “Rough Sets and Flow,” Graphs Lecture Notes in
222-231, 1999. Computer Science, 3641, pp. 1-11, 2005.
[4] Curry, B.; “Rough sets: current and future developments,” Expert [21] Perner, L., F. Alpert and L. Kamins, “How do consumers know which
Systems, 20(5), pp. 247-250, 2003. brand is the market leader or market pioneer? Consumers' inferential
[5] Dimitras, A. I., R. Slowinski , R. Susmaga and C. Zopounidis, processes, confidence and accuracy,” Journal of Marketing
“Business failure prediction using rough sets,” European Journal of Management, 7 (8), pp. 590-611, 2007.
Operational Research, 114 ( 2), pp. 263-280, 1999. [22] Predki, B., R. Slowinski, J. Stefanowski, R. Susmaga and S. Wilk,
[6] Frattasi, S., H. Fathi, F. H. Fitzek, R. Prasad, A. University and M. D. “ROSE software implementation of the rough set theory,” Lecture
Katz, “Defining 4G Technology from the User's Perspective,” IEEE, Notes in Computer Science, 1424, pp. 605-608, 1998.
20 (1), pp. 35-41, January/February 2006. [23] Sabnavis, M. (2002). Getting A fix on the New Middle-Class
[7] Greco, S., Z. Pawlak and R. Slowi´nski, “Generalized decision Consumer, Indian Management, 41 (7), pp. 52-54.
algorithms, rough inference rules and flow graphs,” Rough Sets and [24] Sandhusen, R. L.; Marketing. Hauppauge, New York: Barron’s
Current Trends in Computing, 2475, pp. 951, 2002. Educational Series, Inc., 2000.
[8] Govil, J. and J. Govil, “4G mobile communication systems: Turns, [25] Santhi, K. R., V. K. Srivastava, G. SenthilKumaran and A. Butare,
trends and transition,” International Conference on Convergence “Goals Of True Broad band’s Wireless Next Wave (4G-5G),”
Information Technology, 2007. Vehicular Technology Conference, VTC 2003-Fall, 2003 IEEE 58th, 4
[9] Hodgin, R. C.; “60% of world's population now has cell phone, highest (4), pp. 2317-2321, 2003.
ever”, Retrieved 12/25/09 World Wide [26] Sharma, S. (2004). Understanding urban youth-instant karma. Indian
Web,http://www.tgdaily.com/trendwatch-features/41586-60-of-worlds- Management, 43, (4), pp. 72-81.
population-now-has-cell-phone-highest-ever [27] Shuai, J. J. and H. L. Li, “Using rough set and worst practice DEA in
[10] Ibrahim, J.; “4G Features,” Bechtel telecommunications Technical business failure prediction,” Rough Sets, Fuzzy Sets, Data Mining, and
Journal, 1 (1), pp. 11-14, December 2002. Granular Computing, 3462, pp. 503-510, 2005.
[11] Kotler, P. and Keller, K. L. Identifying Market Segments and Targets, [28] Szimigin, I. and Carrigan, M. . Learning to love the older consumer.
Marketing management, 12th edition, Pearson Education Singapore, p. 2001Journal of Consumer Behavior, 1 (1), pp. 22-34, 2001
234, 2006. [29] Walczak, B. and D. L. Massart, “Rough set theory,” Chemometrics and
[12] Kumar, A. and D. P. Agrawal, “Advertising data analysis using rough Intelligent Laboratory Systems, 47 (1), pp. 1-16, 1999.
sets model,” International Journal of Information Technology and [30] Winer, R. S.; Marketing Management. New Jersey, the United State:
Decision Making, 4 (2), pp.263-276, 2005. Prentice-Hall, Inc., 2000.
[13] Law, R. and N. Au, “Relationship modeling in tourism shopping: a [31] Yao, J. T. and J. P. Herbert, “Financial time-series analysis with rough
decision rules induction approach,” Tourism Management, 21 (3), pp. sets,” Applied Soft Computing, 9 (3), pp. 1000-1007, 2009.

Authorized licensed use limited to: Korea Advanced Inst of Science & Tech - KAIST. Downloaded on March 14,2024 at 05:36:56 UTC from IEEE Xplore. Restrictions apply.