Available online at www.sciencedirect.

com

Expert Systems
with Applications
Expert Systems with Applications 36 (2009) 2513–2524
www.elsevier.com/locate/eswa

Mining tourist imagery to construct destination image position model

Chin-Tsai Lin *, Ya-Ling Huang
Graduate School of Management, Ming Chuan University, 250 Zhong Shan North Road, Section 5, Taipei 111, Taiwan, ROC

Abstract

This study presents a position model for evaluating the image of tourists a destination. The evaluation is based on secondary data
from 1999 through 2004, using a database composed of 20,023 respondents. Data are analyzed using the K-Means data mining method.
Analytical results indicate that the destination image position (DIP) model is established, and four groups of visitor are identified. This
study provides valuable information for tourism planners and marketing professionals attempting to develop offers that are well adapted
to the needs of their target markets.
Ó 2008 Elsevier Ltd. All rights reserved.

Keywords: Data mining; Destination image; Positioning; Formosa

1. Introduction ysis techniques in analyzing the destination image are

factor analysis (Coshall, 2000; Leisen, 2001;), t-testing
Marketing of tourist destinations has recently become (Chaudhary, 2000), perceptual mapping (Kim, 1998),
a very interesting field of research (Castro, Armario, & analysis of means (Baloglu & Brinberg, 1997; Enright
Ruiz, 2007; Enright & Newton, 2004; Hosany, Ekinci, & Newton, 2004; Joppe et al., 2001), cluster analysis
& Uysal, 2006; Pike, 2002). In particular, the image of (Leisen, 2001) and importance performance analysis
a destination has become a very important issue in the (Chaudhary, 2000; Chen, 2001; Enright & Newton,
marketing research in the tourism industry, since many 2004; Joppe et al., 2001). Canonical correlation analysis
countries use promotion and global marketing to support has also recently been applied in destination image
their image and to compete with other destinations. Such research (Hosany et al., 2006). Previous research has
research is increasingly important because the image of a largely concerned the image perception and attitudes
destination is an essential factor in influencing tourist of repeat tourists.
satisfaction and choices (Castro et al., 2007; Han, 1990; Information technology provides many useful tools for
Joppe, Martin, & Waalen, 2001). customers and firms in handling their decision problems
However, image is difficult to determine, because it is (Lin & Hong, 2008). Methods that combine expert systems
highly subjective, including both cognitive (beliefs) and with knowledge management have recently been presented
affective aspects (feelings) (Baloglu & Brinberg, 1997; (Holden, 2004; Li, 2005; Liu & Ong, 2008; Wang & Hong,
Castro et al., 2007; Mackay & Fesenmaier, 2000). 2006). Knowledge management depends on many dat-
Hence, statistical analysis is needed to help firms to abases being available for testing and verification. Addi-
identify the most important factors influencing tourists’ tionally, compares various conventional data analysis
perception, especially for devising marketing strategies. techniques, clearly revealing that knowledge management
Pike (2002) reviewed of studies on destination image yield the most useful analytical results in made practical
from 1973–2000 shows that the most popular data anal- strategy.
Researchers have traditionally applied statistical surveys
*
Corresponding author. Tel.: +886 3 610 2201; fax: +886 3 530 2588. to examine customer behavior. Data mining has recently
E-mail address: ctlin@mail.ypu.edu.tw (C.-T. Lin). been widely adopted to predict customer behavior (Giudici

0957-4174/$ - see front matter Ó 2008 Elsevier Ltd. All rights reserved.
doi:10.1016/j.eswa.2008.01.074
2514 C.-T. Lin, Y.-L. Huang / Expert Systems with Applications 36 (2009) 2513–2524

& Passerone, 2002; Lin & Hong, 2008; Song, Kim, & Kim, ory, and links represent the strength of associations
2001). Data mining is one of the most popular methods for between them. An information node, which can be verbal,
identifying potential customer knowledge in business dat- visual or abstract, is activated in working memory when a
abases to improve decision-making in a firm, by automat- stimulus is present in an external environment (such as a
ically or semi-automatically mining and analyzing large word pronounced or picture presented), or when an inter-
amounts of data in a business database, and discovering nal production occurs (such as when the picture is encoded
potential significant underlying patterns and rules in the and interpreted).
data (Hsu & Chen, 2007; Liao, Hsieh, & Huang, 2008; Fig. 1 shows the proposed framework for integrating the
Lin & Hong, 2008; Wang & Hong, 2006). K-means is internal and external factors identified in previous litera-
widely used to analyze market segmentation and position- ture. A two-dimensional destination image position (DIP)
ing (Kim & Ahn, 2008). Data mining is undoubtedly an model is presented, in which one dimension denotes the
effective way of discovering the image of a destination for internal driving forces, namely place formation processes,
marketing. and the other dimension represents the external driving
Data mining is very effective for determining destination forces, namely the destination image evaluation. The DIP
image marketing strategies. Previous studies have applied model can scale these factors to compare empirical applica-
data mining in tourism marketing research, but none have tion or static scales for a set of similar destinations by the
considered destination image. Therefore, this study has two K-Means method. The positioning process also involves a
principal aims: (1) to adopt data mining to build a position segmentation process, enabling the targeting of a destina-
model for assessing a destination image and (2) to improve tion to the visitors who are most likely to appreciate it
understanding of a destination’s ability to satisfy, provide (Mykletun, Crotts, & Mykletun, 2001). Therefore, this
value and influence tourists to visit the destination again. model provides a new method of categorizing destination
The authors hope that the information provided in this image positions, based on the appeal of a destination to
study will help tourism planners and marketing profession- its visitors.
als to develop offers that fit the requirements of their target
markets. Moreover, the proposed technique could also 3. Method
prove useful in marketing various tourist destinations.
3.1. Data collection
2. Destination image position (DIP) model
This study used secondary data accumulated from Tai-
Individuals associate images with specific places. A des- wan Tourism Bureau Ministry of Transportation and
tination image is defined in terms of the sensual, aesthetic Communication, ROC. (The Tourism Bureau, M.O.T.C.)
and emotional dimensions of space, and is often based on Visitors Expenditure and Trends in Taiwan Survey Report,
contradictory emotions (González, 2005). The sense of which is a yearly statistical report of the Tourism Bureau,
place is created through experience in the spatial setting. M.O.T.C., based on responses from overseas tourists. This
Abstract space, which lacks significance other than strange- work analyses is conduct with data that covered a period
ness, becomes a concrete place filled with meaning through from 2000 through 2005. The total number of observations
human experience (Tuan, 1977). Increasingly, the abstract in the sample was 20,023.
becomes particular; the amorphous becomes concrete; con- Destination image and personal statistics were based
tingencies become commitments, and movement becomes on data filed by the Tourism Bureau, M.O.T.C. Statistics
pause. A destination image is generally in terms of three Executive Information System. The data were collected
sets of criteria, based on (1) regulatory framework, (2) busi- from two international airports. This study summarizes
ness environment & infrastructure and (3) human, cultural destination image, satisfaction, travel trends and personal
and natural resources. The first of these categories denotes details of visitors filed with the Tourism Bureau,
elements that are policy-related and generally under the M.O.T.C. in writing, by questionnaire in person. The fol-
control of the government; the second category represents lowing ten destination image variables are adopted: envi-
elements of the business environment and the economic ronmental sanitation, public security, recreational
infrastructure of, and the third category indicates the facilities, traffic, price, food & beverages, history & cul-
human and cultural elements of a destination’s resources ture, friendliness of people, climate and scenery & land-
(The World Economic Forum, 2007). scape. The respondents were asked to state the extent of
The image of a destination brand can be defined as the their agreement to these image items on a three-point
perceptions of a place as reflected by the associations held scale.
in a tourist’s memory (Cai, 2002). Anderson’s (1983) psy-
chological theory of adaptive control of thought to 3.2. Data analysis
describe that a person’s knowledge structure or memory
as ‘‘a simple network in which all elements or units are Fig. 2 illustrates the stages of data analysis. A two-stage
nodes, and the connections among them are links”. The method was used to build the DIP model. In the first stage,
nodes denote various pieces of information stored in mem- Clementine K-Means was applied to mine the destination
 C.-T. Lin, Y.-L. Huang / Expert Systems with Applications 36 (2009) 2513–2524 2515

Regulatory Dimension Business Dimension Resource Dimension

 Environmental sanitation  Recreational facilities  History and cultural
 Public security  Traffic  Friendliness of people
 Price  Climate
 Food and beverage  Scenery and landscape

External
Forces

Evaluation
Formation
Space Place

Internal
Forces

Collective
in
Memory
Filled
with
Meaning
Field
of
Care
Recognize
and
Experience

Fig. 1. Research model of destination imaging.

Data Collection Stage 1 Stage 2

Cluster 1 Cluster 2
K-Means Mapping
Position
Data Base
Cluster 3 Cluster 4 Model

Fig. 2. The system of data analysis stages.

images of international visitors. Cluster K-Means is opti- ters in a dataset. The second stage adopts the analysis
mized for speed, and is suitable for very large social sur- result of first stage to build and construct the DIP model.
veys. Moreover, K-Means can handle different types of
variables, especially those that occur in survey question- 3.3. Sample distribution
naires and database records. It is developed specifically
for use in market segmentation, and offers several unique The total number of observations in the sample, as
features. Clustering is one of the most popular data mining shown in Table 1, was 20,023. The samples mainly com-
methods. The goal of clustering is to identify distinct clus- prised Japan (39.4%) and American (12.3%) visitors, with
2516 C.-T. Lin, Y.-L. Huang / Expert Systems with Applications 36 (2009) 2513–2524

Table 1 (27.1%). Males constituted 74.1%, while females comprised

Sample distribution 25.9%.
Items n % Items n %
Nationality Occupation 4. Results
Japan 7895 39.4 Legislators and 4,951 25.5
managers 4.1. DIP Model for the evaluation of Formosa destination
Singapore 370 1.8 Professionals 5,183 26.7
Korea 981 4.9 Technicians and 1,676 8.6
image
assistant
Malaysia 141 0.7 Clerks 3,559 18.3 This DIP model is a multi-dimensional diagram com-
Indonesia 97 0.5 Service workers and 872 4.5 prising phases supported by the influential factors of desti-
shop/market salesman nation image. This preliminary model is appropriate for a
Other areas of 1074 5.4 Agricultural, 155 0.8
Asia husbandry, forestry
decision maker or project planner of destination shape,
and fishing workers since each of the place planning projects has its own inte-
America 2470 12.3 Production, plant/ 249 1.3 gration methods. The first stage adopts K-Means to cluster
machine operators and the destination images of Taiwan’s international visitors.
labors The second stage builds and maps the DIP model based
Canada 517 2.6 Housewife 913 4.7
England 996 5.0 Students 1,446 7.4
on the analytical result of the first stage.
France 471 2.4 Retired 411 2.1 Fig. 3 shows a conceptual model for evaluating the desti-
Germany 945 4.7 Total 19,415 100.0 nation images of Taiwan. The DIP model assumes that peo-
Other areas of 1375 6.9 ple form images of places through the level of internal forces.
Europe They are evaluating preference or repulsion by destination
New Zealand 653 3.3
and
image variables. The analysis result is based on the following
Australia ten criteria: environmental sanitation (mean = 1.993), public
Others 360 1.8 security (mean = 2.356), history & culture (mean = 2.442),
country scenery & landscape (mean = 2.357), climate (mean = 2.056),
overseas 1676 8.4 friendliness of people (mean = 2.587), recreational facilities
Chinese
Total 20,021 100.0
(mean = 2.178), price (mean = 2.099), food & beverages
Income (US$/per year) Age (mean = 2.446) and traffic (mean = 1.717). Restated, the
Under 10,000 1856 10.4 Under 19 246 1.2 most influential external forces are human factors, cultural
10,000–14,999 970 5.5 20–29 5,432 27.1 factors, and natural resources.
15,000–29,999 1997 11.2 30–39 5,786 28.9 Fig. 3 shows the four clusters of destination image per-
30,000 39,999 2891 16.3 40–49 4,347 21.7
40,000–69,999 4808 27.1 50–59 3,125 15.6
ception to mapping onto the DIP model. The results herein
70,000–99,999 2997 16.9 60 and Over 1,087 5.4 are obtained by comparing the factors described in the fol-
Over 100,000 2251 12.7 Total 20,023 100.0 lowing section, and should be considered as guides to the
Total 17,770 100.0 perception of Taiwan’s destination image.
Education Gender
Primary 131 0.7 Male 14,845 74.1
school
4.2. Cluster destination image
Junior high 3,160 16.1 Female 5,178 25.9
school Table 2 presents the average results of influential factors
College and 16,359 83.3 Total 20,023 100.0 and the levels of affinity for the destination image of Tai-
upward wan. The destination image perceptions for Taiwan are
Total 19,650 100.0 presented in four clusters. These are recognition of local
Note: n = 20,023. customs, regard for environmental construction, respect
for recreational elements and reverence historical human-
ity, called the ‘‘4R segmentation for destination image per-
ception”. Each cluster has particular preference or
relatively few passengers from Malaysia (0.7%) and Indone- repulsion image factors. Comparing with the position value
sia (0.5%) visiting Taiwan among the participants. Approx- of 4R groups, the mean preferences for the destination
imately 27% of respondents had annual incomes of image of Formosa were 2.52, 2.01, 2.05 and 2.38. They
US$40,000–$69,999; 16.9% earned $70,000–$99,999, and are described as follows.
16.3% earned $30,000–$39,999. Over 80% of respondents
had college degrees or above. Over 25% of the respondents 4.2.1. Define cluster 1: recognizing local customs
were professional, including legislators, administrators, The ‘‘recognizing people custom” group comprises visi-
business executives and managers. Moreover, 18% were tors who are interested in the human, cultural, and natural
employed in the service industry, and only 0.8% was agri- resources dimension factors. Broadly, first cluster contrast
cultural, husbandry, forestry or fishing workers. The major- with others presents higher satisfaction in various fields.
ity of respondents were age 30–39 (28.9%) and 20–29 People in cluster 1 are most satisfied with friendliness of
 C.-T. Lin, Y.-L. Huang / Expert Systems with Applications 36 (2009) 2513–2524 2517

Environmental sanitation

Traffic Public security

Food and beverage History and cultural

Recognize and experience

Price Field of care Scenery and landscape

Filled with meaning

Recreational facilities Collective in memory Climate

Cluster 1 Cluster 2
Friendliness of people Cluster 3 Cluster 4

Fig. 3. Position model for the destination image of Formosa.

4.2.2. Define cluster 2: regard environmental construction

Visitors in this group emphasize the policy factors gener-
Table 2
ally under the control of the government elements. Visitors
The result of K-Means in this second cluster have the lower satisfaction as opposed
Factor Cluster 1 Cluster 2 Cluster 3 Cluster 4
to others, most notably among traffic (mean = 1.068) and
Recognize Regard Respect Reverence environmental sanitation (mean = 1.696) factors. Namely,
people environmental recreational historical cluster 2 visitors regard regulatory framework (policy rules
custom construction element humanity and regulations, environmental regulation, safety and secu-
(n = 4862) (n = 6367) (n = 6333) (n = 2461) rity, and health and hygiene) as more important than either
Mean Mean Mean Mean
the business environment and infrastructure or factors
Environmental 2.308 1.696 2.019 2.072 relating to human, cultural and natural resources. In sum-
sanitation
Public security 2.714 2.206 2.089 2.440
mary, people in this group are unhappy with elements that
Recreational 2.292 1.970 2.016 2.388 are policy-related and generally under the control of the
facilities government. Therefore, this cluster is named the ‘‘regard
Traffic 2.001 1.068 2.129 1.818 environmental construction group”.
Price 2.403 1.973 1.875 2.356
Food and 2.804 2.416 2.115 2.593
beverage 4.2.3. Define cluster 3: respect recreational element
History and 2.567 2.251 2.068 2.661 Visitors in the respect recreational element group favor
cultural the business environment and infrastructure attractions.
Friendliness of 2.882 2.636 2.255 2.674 The third cluster has a lower overall satisfaction in terms
people
Climate 2.534 1.749 1.924 2.235
of food and beverage (mean = 1.875) than other clusters.
Scenery and 2.651 2.185 2.055 2.579 Additionally, such visitors have low satisfaction with the
landscape recreational facilities (mean = 2.016), environmental sani-
Level of 2.516 2.015 2.055 2.382 tation (mean = 2.019) and scenery and landscape
affinity (mean = 2.055) factors, suggesting that visitors in cluster
Note: n = 20,023. 3 respect the business environment and infrastructure.
Thus, members of this group respect the business environ-
ment and the economic infrastructure. Hence, this cluster is
people (mean = 2.882) and food & beverages (mean = 2. called the ‘‘respect recreational element group”.
804). Clearly, the people in this group tend to be satisfied
with the human, cultural and natural resources dimension 4.2.4. Define cluster 4: reverence for historical humanity
factors. Therefore, this cluster is called the ‘‘recognizing The reverence historical humanity group comprises visi-
local customs” group. tors who are interested in the history and cultural resources
2518 C.-T. Lin, Y.-L. Huang / Expert Systems with Applications 36 (2009) 2513–2524

Table 3
The 4R destination image perception cluster descriptions
Variable Cluster 1 Cluster 2 Cluster 3 Cluster 4
recognize people regard environmental respect recreational reverence historical
custom construction element humanity
n % n % n % n %
Nationality
Japan 2334 27.9 1632 37.2 2616 85.9 1313 31.2
Singapore 175 2.1 117 2.7 15 0.5 63 1.5
Korea 550 6.6 242 5.5 50 1.6 139 3.3
Malaysia 61 0.7 43 1.0 8 0.3 29 0.7
Indonesia 46 0.5 24 0.5 5 0.2 22 0.5
Other areas of Asia 528 6.3 94 2.1 38 1.2 414 9.8
America 1347 16.1 474 10.8 79 2.6 570 13.6
Canada 258 3.1 152 3.5 28 0.9 79 1.9
England 493 5.9 177 4.0 23 0.8 303 7.2
France 251 3.0 68 1.5 4 0.1 148 3.5
Germany 556 6.6 115 2.6 23 0.8 251 6.0
Other areas of Europe 723 8.6 167 3.8 29 1.0 456 10.8
New Zealand and Australia 343 4.1 127 2.9 25 0.8 158 3.8
Others country 162 1.9 63 1.4 27 0.9 108 2.6
overseas Chinese 549 6.6 898 20.4 76 2.5 153 3.6
Total 8376 100.0 4393 100.0 3046 100.0 4206 100.0
Occupation
Legislators and managers 0 0.0 446 10.6 303 10.3 4202 99.9
Professionals 4050 50.3 873 20.8 260 8.8 0 0.0
Technicians and assistant 1315 16.3 197 4.7 164 5.6 0 0.0
Clerks 1394 17.3 841 20.0 1324 44.9 0 0.0
Service and salesman 356 4.4 345 8.2 171 5.8 0 0.0
Forestry and fishing, etc. workers 95 1.2 19 0.5 41 1.4 0 0.0
Production, operators and labors 185 2.3 17 0.4 46 1.6 1 0.0
Housewife 12 0.1 698 16.6 203 6.9 0 0.0
Students 479 5.9 692 16.5 275 9.3 0 0.0
Retired 168 2.1 77 1.8 163 5.5 3 0.1
Total 8054 100.0 4205 100.0 2950 100.0 4206 100.0
Income (US$/per year)
Under 9999 537 6.8 917 22.8 365 12.8 37 0.9
10,000–14,999 346 4.4 445 11.1 98 3.4 81 2.0
15,000–29,999 894 11.3 648 16.1 268 9.4 187 4.7
30,000–39,999 1499 19.0 637 15.9 441 15.5 314 7.9
40,000–69,999 2403 30.5 569 14.2 825 29.0 1011 25.4
70,000–99,999 1293 16.4 229 5.7 387 13.6 1088 27.3
Over 100,000 691 8.8 112 2.8 237 8.3 1211 30.4
Total 7890 100.0 4014 100.0 2842 100.0 3980 100.0
Education
Primary school 43 0.5 54 1.3 17 0.6 17 0.4
Junior high school 972 11.8 1061 24.7 785 26.5 342 8.2
College and upward 7217 87.7 3179 74.0 2159 72.9 3804 91.4
Total 8232 100.0 4294 100.0 2961 100.0 4163 100.0
Gender
Male 8377 100.0 0 0.0 2275 74.7 4193 99.7
Female 0 0.0 4393 100.0 771 25.3 14 0.3
Total 8377 100 4393 100 3046 100 4207 100
Age
Under 19 75 0.9 132 3.0 35 1.1 4 0.1
20–29 2383 28.4 1799 41.0 913 30.0 337 8.0
30–39 2850 34.0 1127 25.7 709 23.3 1100 26.1
40–49 1790 21.4 701 16.0 445 14.6 1411 33.5
50–59 940 11.2 487 11.1 586 19.2 1112 26.4
60 and over 339 4.0 147 3.3 358 11.8 243 5.8
Total 8377 100.0 4393 100.0 3046 100.0 4207 100.0
Purpose
Tourism 1467 17.9 2487 58.1 1830 60.9 687 16.4
Business 5712 69.6 785 18.3 844 28.1 3256 77.7
 C.-T. Lin, Y.-L. Huang / Expert Systems with Applications 36 (2009) 2513–2524 2519

Table 3 (continued)
Variable Cluster 1 Cluster 2 Cluster 3 Cluster 4
recognize people regard environmental respect recreational reverence historical
custom construction element humanity
n % n % n % n %
Visit relatives and friends 529 6.4 433 10.1 139 4.6 182 4.3
Conference 334 4.1 388 9.1 175 5.8 57 1.4
Studies or education 169 2.1 186 4.3 18 0.6 6 0.1
Total 8211 100.0 4279 100.0 3006 100.0 4188 100.0
Travel style
TS1 469 5.6 799 18.2 569 18.7 346 8.2
TS2 480 5.7 926 21.1 654 21.5 252 6.0
TS3 964 11.5 431 9.8 375 12.3 467 11.1
TS4 65 0.8 32 0.7 25 0.8 26 0.6
TS5 6398 76.4 2204 50.2 1423 46.7 3115 74.1
Total 8376 100.0 4392 100.0 3046 100.0 4206 100.0
Note 1: TS1 = individual package tour Individual package tour; TS2 = group package tour; TS3 = booking tickets and accommodations through travel
agencies; TS4 = self-arranged tour, with activities arranged by a local travel agency; TS5 = self-arranged tour, without activities by a local travel agency.
Note 2: n = 20,023.

90.0 80.0
cluster 1 cluster 2 cluster 1 cluster 2
cluster 3 cluster 4 cluster 3 cluster 4
80.0 70.0

70.0
60.0

60.0
50.0

50.0
40.0
40.0

30.0
30.0

20.0
20.0

10.0 10.0

0.0 0.0
Tourism Business Visit relatives Conference Studies or TS1 TS2 TS3 TS4 TS5
and friends education
Fig. 5. Travel style of 4R clusters. Note: TS1 = individual package tour;
Fig. 4. Travel purpose of 4R clusters.
TS2 = group package tour; TS3 = booking tickets and accommodations
through travel agencies; TS4 = self-arranged tour, with activities arranged
of a destination. The fourth cluster was more content than by a local travel agency; TS5 = self-arranged tour, without activities by a
other clusters with historic and cultural (mean = 2.661) fac- local travel agency.
tors. Visitors in cluster 4 showed strong satisfaction with
friendliness of people (mean = 2.674), food and beverages
(mean = 2.593 and scenery & landscape (mean = 2.579) examine and describe the relationships with the 4R cluster
factors. In other words, they are interested in history and groups and personal responses. The results of the four clus-
cultural resources. Accordingly, members of this group ters are given below.
member prefer the human and cultural elements of Tai- Table 3 shows the demographic segmentation results.
wan’s resources. Therefore, this cluster is named the ‘‘rever- The largest proportion of visitors in all clusters came from
ence historical humanity group”. Japan. The proportion of visitors from Japan was highest
in cluster 3 has, at 85.9%. Almost 100% of tourists in clus-
ter 4 were legislators and managers. Visitors in cluster 4
4.3. Demographic profile of 4R clusters for image segments had the highest average incomes. The largest proportion
tourists with high education levels were found in Clusters
To distinguish tourist data according to the clustering 1 (87.7%) and 4 (91.4%). Gender makes significant contri-
results, a Clementine Relationship Matrix was adopted to butions to the explained variance in these 4R clusters.
2520 C.-T. Lin, Y.-L. Huang / Expert Systems with Applications 36 (2009) 2513–2524

90.0 35.0
cluster 1 cluster 2 cluster 1
80.0 cluster 3 cluster 4 cluster 2
30.0 cluster 3
70.0
cluster 4
60.0
25.0
50.0

40.0 20.0

30.0

20.0 15.0

10.0
10.0
0.0
ny

Ch y
s o ia

e
sia
e

sia

ov rs c ia

e
a

d
an

ea

A pe
da

es
or

ic

tr
an
s

l
an
or

ro
na
p

Ze rea rma

th stra

n
ne
ay

fA

er

in
ap

gl
Ja

ou
K

Eu
r
Ca
m
do
al

F
ng

En

5.0
u
e
A
M

G
In

of
Si

as
ea

e
s

se
an
ar

er
O
er

d
ra

an
th

e
O

al
th
O

0.0
ew

Under 10,000- 15,000- 30,000- 40,000- 70,000- Over

N

9,999 14,999 29,999 39,999 69,999 99,999 100,000

Fig. 6. Nationality of 4R clusters.
Fig. 8. Income of 4R clusters.

Furthermore, the age distributions in each cluster are dif-

results indicate that demographic variables can effectively
ferent. The average age of tourist is oldest in cluster 4.
distinguish variations in the tourist market. Figs. 4–11
The purpose of most passengers in Clusters 2 (58.1%)
show the percentage preference ratings by personal vari-
and 3 (60.9%) was tourism, where are that of most passen-
able in terms of purpose of travel, style of travel, nation-
gers in Clusters 1 (69.6%) and 4 (77.7%) was business.
ality, occupation, income, education, gender and age.
Finally, the type of trip arrangements was different in each
The characteristics of each segment are described as
cluster. Most visitors for tourism preferred individual or
follows.
group package tours. Moreover, most participants pre-
ferred TS5 (self-arranged tour, with activities arranged by
a local travel agency) over TS4 (self-arranged tour, without 4.3.1. Targeting the recognize people custom group
activities by a local travel agency). Figs. 4–11 illustrate the features of segment 1 ‘‘Recog-
Nationality, occupation, income, education, gender, nize People Custom” Group relative to other groups.
age, purpose of visit and travel style each contributed to Segment 1 shows that passengers in this group have ‘‘pur-
the explained variance in these 4R clusters. Analytical pose of travel for business”, ‘‘style of travel prefer TS5

100.0
cluster 1 cluster 2
90.0 cluster 3 cluster 4
80.0

70.0

60.0

50.0

40.0

30.0

20.0

10.0

0.0
s

rs
nt

an

ed
ls

ts
ife
er

er
rk

bo
na

en
ta

sm

tir
ag

rk
le

w
sis
sio

la

ud

Re
C

se
an

le

w
as

nd

St
ou
sa
es
m

.
nd

sa
of

tc

H
d
d

,e
an
an

Pr

sa

or
ng

at
ce
rs

an

er
hi
to

vi
ci

op
fis
sl a

er
ni

S
ch
gi

n,
d
an
Le

ti o
Te

uc
str

od
re

Pr
Fo

Fig. 7. Occupation of 4R clusters.

 C.-T. Lin, Y.-L. Huang / Expert Systems with Applications 36 (2009) 2513–2524 2521

100.0 45.0
cluster 1 cluster 1
90.0 cluster 2 40.0 cluster 2
cluster 3 cluster 3
80.0
cluster 4 cluster 4
35.0

70.0 30.0

60.0
25.0

50.0
20.0

40.0
15.0

30.0
10.0

20.0
5.0

10.0
0.0
Under 19 20-29 30-39 40-49 50-59 60 and over
0.0
Primary school Junior high school College and upward
Fig. 11. Age of 4R clusters.
Fig. 9. Education of 4R clusters.

100.0
cluster 1
90.0 cluster 2
cluster 3
80.0 cluster 4

70.0

60.0

50.0

40.0

30.0

20.0

10.0
Fig. 12. Sun Moon Lake image.
0.0
Man Woman

Fig. 10. Gender of 4R clusters. package tour) relatively higher than other segments”, ‘‘over-
seas Chinese”, ‘‘housewives and students”, ‘‘annual income
is low about US$10,000–$30,000”, ‘‘junior high school”,
(self-arranged tour, without activities by a local travel ‘‘woman” and ‘‘young people under 19 and around 20–29”.
agency)”, ‘‘American”, ‘‘professionals, technicians and
assistant”, ‘‘middle and high income about US$30,000– 4.3.3. Targeting the respect recreational element group
$100,000 per year”, ‘‘college and upward”, ‘‘man” and Segment 3 in Figs. 4–11 shows the features of ‘‘Respect
‘‘middle-aged person aged 30–39” in their demographic Recreational Element Group” contrast with other groups.
profile variables. The typical profile of passengers in this segment is ‘‘pur-
pose of travel for tourism”, ‘‘style of travel are TS1 (indi-
4.3.2. Targeting the regard environmental construction group vidual package tour), TS2 (group package tour), and TS3
Segment 2 in Figs. 4–11 (shows OR illustrates OR depicts (booking tickets and accommodations through travel agen-
OR displays OR presents) the features of segment 2 cies) relatively higher than other groups”, ‘‘overwhelming
‘‘Regard Environmental Construction Group” compared majority is Japanese”, ‘‘clerks”, ‘‘middle and low income
with other groups. The typical profile of passengers in seg- about US$30,000–$70,000 per year”, ‘‘junior high school”,
ment 2 is as follows: ‘‘purpose of travel for tourism”, ‘‘style ‘‘make”, ‘‘young people aged 20–29 second only to segment
of travel are TS1 (individual package tour) and TS2 (group 2” and ‘‘elderly tourists 60 and older”.
2522 C.-T. Lin, Y.-L. Huang / Expert Systems with Applications 36 (2009) 2513–2524

4.3.4. Targeting the reverence historical humanity group Table 4

Segment 4 in Figs. 4–11 shows the features of the ‘‘Rev- The empirical application for Sun Moon Lake
erence Historical Humanity Group”. The typical profile of Factors Advantage Target Characteristic Marketing
respondents in segment 4 is as follows: ‘‘purpose of travel strategy
for business‘‘, ‘‘style of travel prefer TS5 (self-arranged Environmental Recognize Prefer Intensify
tour, without activities by a local travel agency)”, ‘‘an above sanitation people human, promotional
Public security custom cultural, and media, such
average proportion of the respondents reside in other areas Recreational group natural as website,
of Asia, England and other areas of Europe than other facilities resource. magazine and
groups”, ‘‘nearly all legislators and mangers”, ‘‘high income Traffic Japanese, guidebook.
about US$40,000–$100,000 per year”, ‘‘college and Price American Design a
upward”, ‘‘man” and ‘‘people in middle age about 40–59”. Food and and simple and
beverage European. short-term
History and High trip Make
cultural education, partnership
4.4. Empirical application Friendliness of middle range with business
people income, man hotels or
This study adopted the Sun Moon Lake of the famous Climate and middle- airline
destination of Taiwan for the DIP model empirical applica- Scenery and aged.Tend to companies
landscape self-arranged
tion examples. The DIP model and 4R segmentation for tour, without
destination image perception are applied as follows: activities by a
local travel
Step 1: Build and Identify the well regarded resources of agency
the destination, and map them onto the DIP model. Note: = advantage images; = adequate images; = weakness
Step 2: Look for clusters in the DIP model. images.
Step 3: Carry on contrasting and comparing the demo-
graphic profile based on the segment result of this
research.
Step 4: Segment the target consumer through the character- prefer this region, since they are interested in the human,
istic of cluster, then draft the marketing strategy of cultural and natural resources.
attractive and customization in view of the above.
4.4.1.3. Step 3: Analyze the demographic profile. As
explained in step 2, Sun Moon Lake was paired with the
4.4.1. Destination: Sun Moon Lake recognize people custom group, and then the target cus-
4.4.1.1. Step 1: Identify the destination. Fig. 12 presents the tomer characteristics were analyzed. The respondents in
imagery of the Sun Moon Lake, which is located in the cluster 1 are geographically diverse. An above average pro-
middle of Taiwan at an elevation of 748 m above sea level, portion of the respondents reside in Japan, America and
is the only large natural lake in Taiwan. The southern part Europe. The cluster members are mainly professionals,
of Kuang–Hua Island is shaped like a new moon, and the high education, middle range income, male and middle-
northern part is shaped like a sun; hence the name Sun aged. An above average proportion of the cluster members
Moon Lake. The Sun Moon Lake is a immense stretch of are the 30–39 year age bracket. The main purpose for travel
deep blue-green water, which looked like in earlier times is business. Furthermore, the members tend to do self-
when it was a shallow marsh covered with vibrant green arranged tours, without activities by a local travel agency.
water plants. The natural forests bordering these roads
are good places for bird watching. Moreover, other activi- 4.4.1.4. Step 4: Devise marketing strategies. Table 4
ties available around the Sun Moon lake include rowing, shows that the Sun Moon Lake suits individuals who
taking cruise ship, walking nature trails and riding bicycle wish to experience a destination’s history, cultural, local
activities to encircle the. The first known tribe to live in the customs and practices and meet local residents. Regard
Sun Moon Lake region was the Shao Clan. The Harvest America, Europe and Japan as and promote the country
Festival, Sowing Festival and the special handicraft fair mainly. Furthermore, segment 1 members prefer self-
held every year, as well as their articles has helped to pre- arranged tours, so planner/operator should intensify pro-
serve the particular culture of the Sun Moon Lake region. motional media, such as websites, magazines and guide-
The clustered village of Thao indigenous people are attrac- book, in order to increase opportunity touching the
tive and unique features. target market. Moreover, business travelers generally stay
for a short time at their destination. The tourism propri-
4.4.1.2. Step 2: Match the cluster. Step 1 was performed to etor can design a simple and short-term trip, and than
identify the Sun Moon Lake’s advantages and weakness. form partnerships with business hotels or airline compa-
Local culture is strong in this area. Clearly, visitors in clus- nies to increase visibility and chance of being chosen by
ter 1, the recognize people custom group, are most likely to potential customers.
 C.-T. Lin, Y.-L. Huang / Expert Systems with Applications 36 (2009) 2513–2524 2523

5. Conclusions particular images to attract different segments. This

approach of targeting the most approachable target mar-
The DIP model provides a highly effective evaluation kets is likely to enhance the efficiency and effectiveness of
tool for planners, managers and government officials to the destination’s promotional revenue.
help make decision for marketing strategies, particularly Most importantly, this study builds a position model for
for segmenting and positioning. The model can be applied assessing the image of a destination. This DIP model can
ideally to posit all types of destination (e.g. a theme park, help firms to construct sustainable competitive advantage,
city or county) onto this model. The main contributions and to create competitive advantage for an entire area
of this study are the use of k-means to construct the 4R through positioning and segmenting.
evaluation model, and the evaluation of the insightful pat-
terns of destination image by combining destination image Acknowledgment
factors and place formation theories. This study explores
the destination’s internal and external formation on the The authors would like to thank the Tourism Bureau Min-
levels of DIP evaluation. Various strategic and evaluative istry of Transportation and Communication of the Republic
position models have been developed in order to attach of China, Taiwan, for database supporting this research.
those strategically in regard to destination image theories,
and previous research has identified many criteria for des-
tination image evaluation. References
The increase in market-driven competition and adoption
Anderson, J. (1983). The Architecture of Cognition. Cambridge, MA:
of information technology causes tourism market conditions Harvard University Press.
and visitors’ needs to change rapidly. Travelers are no longer Baloglu, S., & Brinberg, D. (1997). Affective images of tourism destina-
satisfied with mass tourism producing products anymore, tion. Journal of Travel Research, 35, 11–15.
and mass tourism customization becomes an alternative to Cai, L. A. (2002). Cooperative branding for rural destinations. Annals of
Tourism Research, 29, 720–742.
destinations producing variety a products that are limited
Castro, C. B., Armario, E. M., & Ruiz, D. M. (2007). The influence of
in resource and reasonable cost structure. The ability to ful- market heterogeneity on the relationship between a destination’s image
fill diverse tourists’ needs is insufficient to catch the new and tourists’ future behaviour. Tourism Management, 28, 175–187.
opportunities if destinations cannot sense the changes. A Chaudhary, M. (2000). India’s image as a tourist destination: A
destination’s performance on mass tour customization capa- perspective of foreign tourists. Tourism Management, 21, 293–297.
Chen, J. S. (2001). A case study of Korean outbound travelers’ destination
bility and agility all depend on its demarcation capabilities.
images by using correspondence analysis. Tourism Management, 22,
This study applies data mining to analyze an enormous 345–350.
data set, and helps define the customization and contribu- Coshall, J. T. (2000). Measurement of tourists’ images: The repertory grid
tion, thus formulating the DIP model application approach. Journal of Travel Research, 39, 85–89.
approach. Segments of target markets are identified for Enright, M. J., & Newton, J. (2004). Tourism destination competitiveness:
A quantitative approach. Tourism Management, 25, 777–788.
defining marketing strategies. The results indicate the pref-
Giudici, P., & Passerone, G. (2002). Data mining of association structures
erences and repulsion imagery features through market to model consumer behavior. Computational Statistics and Data
segments. The groups with the satisfied imagery are clearly Analysis, 38, 533–541.
the striking segments. These segments form the most sensi- González, B. M. (2005). Topophilia and Topophobia: The home as an
tive target markets, as they not only have a preference evocative place of contradictory emotions. Space and Culture, 8, 193–213.
Han, C. M. (1990). Testing the role of country image in consumer choice
image but also have the strongest visitation tendencies.
behaviour. European Journal of Marketing, 24, 24–40.
Additionally, the demographic characteristics of the seg- Holden, N. (2004). Why marketers need a new concept of culture for the
ment members can be used to designate promotion media global knowledge economy. International Marketing Review, 21,
with matching visitor profiles. 563–572.
The 4R segmentation for destination image perception Hosany, S., Ekinci, Y., & Uysal, M. (2006). Destination image and
destination personality: An application of branding theories to tourism
is clearly an easy and simple way to describe and seg-
places. Journal of Business Research, 59, 638–642.
ment tourist’s features. Hence, 4R segmentation consid- Hsu, C. C., & Chen, Y. C. (2007). Mining of mixed data with application
ers four classes of tourist in the tourism market. to catalog marketing. Expert Systems with Applications, 32, 12–23.
Moreover, a clustered destination image is produced. Joppe, M., Martin, D. W., & Waalen, J. (2001). Toronto’s image as a
The clusters are favorable to market segmentation and destination: A comparative importance-satisfaction analysis by origin
of visitors. Journal of Travel Research, 39, 252–260.
creating marketing strategies. From a competitive per-
Kim, H. B. (1998). Perceived attractiveness of Korean destinations. Annals
spective, destination operators/marketers can further of Tourism Research, 25, 340–361.
enhance their image by directing promotional campaigns Kim, K. J., & Ahn, H. (2008). A recommender system using GA K-Means
towards specific markets, increasing the popularity of a clustering in an online shopping market. Expert Systems with
destination over other areas that a traveler might con- Applications, 34, 1200–1209.
Leisen, B. (2001). Image segmentation: The case of a tourism destination.
sider when selecting a destination. This investigation
Journal of Services Marketing, 15, 49–66.
demonstrates that the profiles of customers in each seg- Li, S. L. (2005). A web-enabled hybrid approach to strategic marketing
ment are important factors in determining destination planning: Group Delphi plus a web-based expert system. Expert
image. This finding enables decision makers to shape Systems with Applications, 29, 393–400.
2524 C.-T. Lin, Y.-L. Huang / Expert Systems with Applications 36 (2009) 2513–2524

Liao, S. H., Hsieh, C. L., & Huang, S. P. (2008). Mining product maps for Pike, S. (2002). Destination image analysis-a review of 142 papers from
new product development. Expert Systems with Applications, 34, 50–62. 1973 to 2000. Tourism Management, 23, 541–549.
Lin, C., & Hong, C. (2008). Using customer knowledge in designing Song, H. S., Kim, J. K., & Kim, S. H. (2001). Mining the change of
electronic catalog. Expert Systems with Applications, 34, 119–127. customer behavior in an internet shopping mall. Expert Systems with
Liu, H. H., & Ong, C. S. (2008). Variable selection in clustering for Applications, 21, 157–168.
marketing segmentation using genetic algorithms. Expert Systems with The World Economic Forum. (2007). The travel & tourism competitiveness
Applications, 34, 502–510. report 2007. Geneva, Switzerland.
MacKay, K. J., & Fesenmaier, D. R. (2000). An exploration of cross-cultural Tuan, Y. F. (1977). Space and place: The perspective of experience.
destination image assessment. Journal of Travel Research, 38, 417–423. London: Edward Arnold.
Mykletun, R. J., Crotts, J. C., & Mykletun, A. (2001). Positioning an Wang, H. F., & Hong, W. K. (2006). Managing customer profitability in a
island destination in the peripheral area of the Baltics: A flexible competitive market by continuous data mining. Industrial Marketing
approach to market segmentation. Tourism Management, 22, 493–500. Management, 35, 715–723.