48 International Journal of Asian Business and Information Management, 5(4), 48-59, October-December 2014

Knowledge Discovery for

Tourism Using Data Mining
and Qualitative Analysis:
A Case Study at Johor Bahru, Malaysia
Atae Rezaei Aghdam, Department of Information Systems, Universiti Teknologi Malaysia,
Johor Bahru, Malaysia
Mostafa Kamalpour, Faculty of Computing, Universiti Teknologi Malaysia, Johor Bahru,
Malaysia
Alex Tze Hiang Sim, Department of Information System, Universiti Teknologi Malaysia, Johor
Bahru, Malaysia

ABSTRACT
This paper aims to propose a new guideline for analyzing tourist profiles as found in www.tripadvisor.com.
These have been examined from two different aspects so as to gain conclusive results. Tourist data were
“crawled” from tripadvisor.com through a specific web crawler. Mining techniques using a combination of
visualization, clustering, and association rules were instrumental in discovering the first set of interesting
knowledge. This was followed by a qualitative analysis applied through Nvivo software via coding of the tour-
ist’s comments in order to define the design of the prospective model. A final set of results was obtained once
both results confirmed each other. In this study, results show that there are several types of tourists; with each
group having different preferences. For example: male Singaporean visitors to hotels tend to enjoy wine and
food in addition to outdoor activities; while local visitors to Legoland are not satisfied with certain aspects,
such as the price of food. International tourists, however, consider the affirmative points of Legoland. This
research can be very useful for tourist associations and hotel managers in Johor Bahru.

Keywords: Data Mining, Johor Bahru, Knowledge Discovery, Malaysia, Qualitative Analysis, Tourism

1. INTRODUCTION tourists arriving in Malaysia. In addition, the

According to the annual report of Tourism Ma-
laysia in 2010 (Tourism Malaysia, 2010), the
tourism industry has played an important role
in increasing the GDP with some 24.6 million
World Tourism Organization estimates that
global tourism is expected to grow faster than
other economic sectors in the world (Witten
& Frank 2011). Uncovering new, interesting
and useful information on tourist data can be

DOI: 10.4018/ijabim.2014100105

Copyright © 2014, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
 International Journal of Asian Business and Information Management, 5(4), 48-59, October-December 2014 49
helpful for tourism organizations in order to
identify tourists’ behavior patterns and their
preferences. This, then, is a great opportunity
for the Malaysian tourism industry to boost
tourist arrivals and increase revenue. This paper
focuses on the role of data mining in the Malay-
sian tourism industry, in particular, regarding
places of interest in Johor Bahru and hotels in
the Mersing area. In essence, we will analyze
tourist behavior patterns in Johor Bahru so as
to discover useful and hidden knowledge in
order to recommend appropriate places to visit.
Moreover, quantitative and qualitative analyses
have been undertaken simultaneously in this
paper. Quantitative analysis was applied by
Weka machine learning software; while qualita-
tive analysis was performed by Nvivo software
based on tourists’ comments. Each group of
analysis findings supports the other in order to
obtain significant results. This study can assist
tourist associations in Johor Bahru and travel
agencies in the promotion of places in Johor
Bahru which would be attractive to both local
and international visitors. This paper is arranged
in the following manner: Section 1; investiga-
tion of related works regarding data mining in
the tourism industry; Section 2; explanation of
our research methodology and pre-processing
procedure; and finally, Section 3; analysis of the
data and illustration of “hidden knowledge” in
order to recommend these to tourists.
2. LITERATURE REVIEW
Currently, data mining supports various kinds
of application tasks: from data pre-processing
to association rules discovery, data clas-
sification, and cluster analysis respectively
(Witten & Frank, 2011). Actually, it is part of
the decision-making process; and the avail-
ability to analyze data automatically helps to
determine a potential model. It also assists in
estimating customer behavior in the realm of
enhancing decision makers’ ability to both
adjust marketing strategy and reduce risks
(Li, 2012; Han & Kamber 2006). In another
study (Bose, 2009), concentration was given to
Copyright © 2014, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
three main aspects of using data mining in the
tourism industry. These are, namely; forecast-
ing tourist expenditure, analyzing profiles of
tourists and forecasting the number of tourist
arrivals. The author has found various results
based on these three dimensions. For instance,
in forecasting tourist expenditure, artificial
intelligence sources such as Neural Network
were used for estimating tourist expenditure in
the Balearic Islands. Further, Au & Law (2002)
used data mining techniques to predict shop-
ping expenditure by tourists with an accuracy
level of 94%. In relation to analyzing tourist
profiles (Bose, 2009), categorized tourists into
specific groups such as; developmental support,
prudent developers, ambivalent, cautious and
protectionist respectively by using a clustering
technique. In forecasting tourist arrivals, some
studies had been conducted examining tourist
arrivals to Hong Kong from six different coun-
tries; Artificial Neural Network (ANN) showed
that this outweighed statistical methods. In a
further study, Bose (2009) stated that it has been
argued that, to date, only some AI techniques
such as ANN and clustering techniques have
been used in tourism data mining. It is largely
prepackaged software that uses these techniques
readily; it can also be used with little training
to analyze data. However, the author believed
that, in the future, more than one method will
be applied for analyzing data. In this paper we
focus on analyzing tourist profiles from two
different perspectives, namely: quantitative
and qualitative. In the following sections, the
related studies illustrate how traditional data
mining techniques have been used in the tour-
ism industry.
2.1. General Research into
the Tourism Industry
Data mining offers different approaches by
which to gain knowledge and subsequently use
this knowledge in the realm of, specifically:
enhancing revenue, identifying customers’ be-
havior patterns, recommendations to customers,
etc. In the tourism industry in particular, a great
deal of work has been devoted so far, but still
50 International Journal of Asian Business and Information Management, 5(4), 48-59, October-December 2014
the domain of research in this area is vast. More
research studies are needed in order to provide
better results compared to what we have now.
For instance, using data mining techniques
based on the information as mentioned can
assist hotels to stay competitive. Hotels can
identify the most profitable methods by which
to develop a customer loyalty program. For
example, tourism revenue in South Korea has
doubled in the last decade. The reason for this
is that South Korea expanded their capacity
for hosting visitors by renewing their hotels
and increasing the number of hotel rooms.
Subsequently, competition between hoteliers
increased. A survey of 281 customers helped to
edge closer to the answers of some questions,
such as: which customers are likely to return
to the same hotel; or, which service attribute
is more important to customers and so on. The
survey took place among 281 guests in 11
hotels in Seoul, South Korea. Many character-
istic information samples were collected and
considered as sources of data; these included
price range, location and service amenities
respectively. Visitors also provided the authors
with data related to their demographic profile,
including: age, occupation, gender, nationality
and so on. All participants had visited at least
one of the 11 hotels (Kosala, 2000). A study by
Poon (2005) researched the level of satisfac-
tion with Malaysian hotels among travelers.
Their research question aimed to find differ-
ences between Asian and Western travelers in
an attempt to evaluate customer satisfaction
issues with Malaysian hotels. The goal of the
study was to suggest some ways to improve
hotel services, as well as examining outlying
factors that provide specific services, which
may encourage visitors to revisit the country.
By distributing a questionnaire at KLIA Air-
port among 200 tourists who were leaving the
country, data was collected and analyzed. The
results showed that 43% of visitors were in the
age group of 21-35 and 32% of them were in the
age group of 36-50. There are some attributes
in the questionnaire relating to factors such as:
cleanliness, rooms, location, sleep quality, value
for money and services. Results show that two
Copyright © 2014, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
of these attributes are extremely important for
travelers, specifically: price (value), and food.
The attribute value is highly associated with
customer loyalty. The first result shows that
Asian travelers pay more attention to the prices
offered by hotels as compared with Western
travelers. The second result shows that, having
good food and hospitality is important to both
Western and Asian groups. Similar research
has been carried out by Choi (2000) regarding
Asian and Western travelers in Hong Kong. The
method of data collection in this research was by
use of a questionnaire in the Hong Kong airport
and distributed among those visitors who live in
the country. In the questionnaire, travelers were
asked questions relating to demographic and
personal issues, as well as questions relating to
the hotels where visitors resided. The feedback
regarding hotels included: cleanliness, location,
room rate, service, value for money and location.
540 people responded to the questionnaire and
the collected data was analyzed by using factor
analysis. Results showed that Asian travelers
generally spend less money than Westerners.
One of the reasons for this issue is that most
Asian countries are among developing nations;
as a result, their salary is normally less when
compared with Westerners. Analysis also shows
that Westerners spend approximately 45% of
their budget on accommodation, whereas Asians
spend less than 25% on this. Another result,
which supports the previous one, is that 70%
of Asians look for midrange hotels. Results
also identified that Asians are more interested
in shopping than Westerners. Asians generally
like to spend more than 50% of their budget on
shopping. They also tend to place more emphasis
on value for money, whereas Westerners are
concerned with room quality. The final result
shows that Asians are very interested in spend-
ing their time on entertainment during the trip.
2.2. Data Mining in
Tourism Industry
Discovering new, stimulating and valuable
knowledge through the use of a variety of
techniques such as; classification, clustering,
 International Journal of Asian Business and Information Management, 5(4), 48-59, October-December 2014 51
visualization and association rules is the main
aim of data mining (Romero, 2009). In the
meantime, the necessity of using data mining
in tourism is inevitable. In this section, we will
investigate some related works regarding data
mining in the tourism industry. For instance,
Guoxia (2009) focused on using a decision
tree algorithm to analyze tourist markets from
two different dimensions including: impact
factors of tourist spending and impact factors
of tourists’ comprehensive evaluation of tourist
destinations respectively. The first step focuses
on total expenditure of tourists in the classifica-
tion objects in the decision tree model; while
the second step focuses on attributes and values
including: cost, traffic, shopping environment,
dining etc. By using the decision tree algorithm
and associative rules mining to analyze the
data and some rules, it was found that, for
example, poor shopping environments affect
tourist sentiment even if they are located in an
area having the best attractions. Another result
illustrates that, if tourism products only focused
on an attraction’s landscape in itself, the tour-
ism product would certainly decrease the value
itself. Another study proposed a new semantic
association rule mining algorithm, which intro-
duced a genetic algorithm. This method deals
with textual information and divided character-
istic words into various categories in an attempt
to find association rules between them. They
categorized tourist emergency information into
five categories, namely: object, environment,
activity, event and result. The author believes
that this technique is useful for searching rules
in a database. The results indicate, further,
that the algorithm can obtain semantic rules
effectively with a higher extraction rate and
correctness. In extracted association rules with
high confidence, the appearing words present
the concepts of an emergency event. Hence,
several rules discovered from small associa-
tion word sets have been used to search for
emergency reports of tourism. Furthermore,
semantic association rules mining would obtain
a better result from a more sophisticated text
set of tourism emergency information (Zhou,
2008). Data mining is present in occasions of
Copyright © 2014, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
every part of tourism such as the hotel industry;
for instance, by offering tourist information
kiosks to provide customer information dur-
ing their stay and then analyzing their data. A
research study was carried out in South Korea
regarding data mining; the hotel industry pro-
filed visitors based on their preferences and
comments. The author attempted to examine
the effect of a customer’s demographic profile
on hotel choice. Visitors were categorized by
some factors including: country of origin, oc-
cupation, sex, age etc; as a result, they identified
interesting relationships existing between items
in a dataset. Finally they came up with some
“if-then” rules, which show the probability of
existence of some facts by analyzing the data.
For instance, one of the rules indicates that if
a customer stays at the hotel for a convention,
then he/she is probably a manager. From this
finding that the customer is a manager, we can
acquire other information based on findings
about managers. For example, if a customer
prefers either a room with a single bed or a suite,
pays for the hotel room by credit card, and stays
at the hotel for travel/pleasure; the customer is
likely to be Korean. Further, the author found
that some combinations of customer character-
istics have more influence on the customer’s
hotel selection and patronage behavior than a
single attribute. Hotel management would be
wise to take into account a multitude of attributes
such as: customer’s demographic profile, travel
purpose, prior service experiences, as well as
the availability of certain amenities (Min Hokey,
2002). It was just one of the examples among
the rules that they had defined.
2.3. Techniques in Tourism Data
The most important unsupervised learning tech-
nique is that of clustering. Accordingly, attention
is focused on ways to identify a structure in a
group or collection of unlabeled data in order
to organize elements into groups (Isa, 2009).
Clustering refers to a process of clustering
and grouping together a collection of abstract
or physical objects into similar groups. In a
cluster, objects of one group are different from
52 International Journal of Asian Business and Information Management, 5(4), 48-59, October-December 2014
the objects of other groups (Etemadi, 2010).
In this paper, we intend to discover interest-
ing knowledge which can form the enormous
amount of data in the web by using the clus-
tering technique. A cluster can be known as a
collection of data elements, which are similar to
each other and also different from other cluster
elements. Clustering refers to grouping data
into clusters with minimum similarity existing
between elements of two different clusters and
maximum similarity between elements of each
cluster (Larose, 2005). Another useful technique
for discovering rules between elements is by
Association Rule Mining. This method is a data
mining technique, which was introduced by
Agrawal in 1993. He proposed the algorithm
of Apriori in 1994 (Agrawal, 1996). The main
idea of Apriori is to find a relationship between
different items of a database. The association
technique is useful to find a relationship be-
tween two or more attributes and it attempts to
find rules between attributes. Association rules
mining helps to extract this kind of knowledge
from a database so that it can be used by recom-
mendation systems.
3. METHODOLOGY
Our methodology is categorized into five phases
which include: (3.1.) defining required data,
(3.2.) data collections, (3.3.) pre-processing,
(3.4.) data mining and (3.5.) conclusion.
3.1. Defining Required Data
The phase relating to defining required data
focuses on distinguishing which type of data
needs to be analysed. Thus, the first step is
detection of the tourist data. For identifying
required data, we found a target website which
gathers significant data about tourists, including:
age, gender, country of origin, goal of travel,
travel partners etc. Therefore, www.tripadvisor.
com was chosen as a repository within which
to collect the data. Tripadvisor is the world’s
largest travel website, with more than 200
million unique monthly visitors and over 100
million reviews and opinions. In addition, it has
Copyright © 2014, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
a comprehensive storage of tourist data which
could be used in our research.
3.2. Data Collections
Due to the formidable task of gathering data
manually, the data were crawled from www.tri-
padvisor.com by providing the task specific web
crawler, which was named as “RK Crawler”.
RK crawler is able to obtain a special place
name, search for it in the tripadvisor database,
and then extract the tourist’s data concerning
this place.
After collecting data by means of the
crawler, data will then be exported automatically
to Microsoft Excel with CSV format.
The extracted dataset includes various
attributes such as: age, gender, country, rat-
ing, who the tourist traveled with, purpose of
travel, tourist type, tourist’s contribution, visited
cities, title, comments, places of interest, etc.
Visualizing some significant attribute of data
can demonstrate relationships between items.
The results show that, if a visitor is interested
in spas, there is a 67% probability that he/she
is also interested in going shopping.
Our results show that if a visitor is inter-
ested in theme and amusement parks, there
is a 73% chance that he/she is interested in
shopping as well.
3.3. Pre-Processing
Pre-processing is the third step of our methodol-
ogy. In this phase, the data needs to be clean,
normalized and transformed. Further, the col-
lected data needs to be compatible with a suitable
format to use in Weka data mining software.
For preparing data with ARFF format, we need
to follow some rules regarding standardization
of this format to load in Weka software. Figure
1 demonstrates a collection of data with ARFF
format for loading in Weka software.
3.4. Data Mining
The most important part of this research is
mining the data. Unlike traditional data mining
techniques, our research includes quantitative
 International Journal of Asian Business and Information Management, 5(4), 48-59, October-December 2014 53
Figure 1. Weka tags by ARFF format
and qualitative analysis so as to obtain conclu-
sive results. We selected Weka software to apply
visualization, clustering, and association rules
mining to attempt to find stimulating knowl-
edge and rules existing between elements in a
dataset. We focused on two main techniques,
namely: clustering and association. Clustering
places an emphasis on building models, which
are able to assign new instances to one of a set
of well-defined classes. By using clustering
techniques, we can determine the segmentation
of customers. K-Means and EM algorithm were
applied for clustering the data. Meanwhile, an
Apriori algorithm was applied to discover the
association rules existing between items in our
dataset. Likewise, for acquiring more detailed
understanding and extraction of knowledge
from numeric and non-numeric information
forms, Nvivo software is highly recommended
(Bazeley, 2013; Richards, 1999). Qualitative
and text analysis is then fulfilled by Nvivo
through coding of the tourist’s comments in
order to draw out the model. The model can
be delivered via significant points of places
Copyright © 2014, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
and tourists’ preferences. Therefore, the results
from the Nvivo part will support the findings
of the Weka section.
3.5. Conclusion
Conclusion is the last phase of our methodol-
ogy. In this phase, interesting knowledge will
be revealed from the data; also, outcomes of
Weka and Nvivo will be compared with each
other to gain conclusive results. According to
results, the places would then be recommended
to specific groups of tourists based on their
preferences. On the other hand, the output of
this phase is the output of entire phases of this
methodology.
4. EXPERIMENTAL RESULTS
In this paper, several analyses have been car-
ried out on data from different dimensions. We
analyzed the data of 587 profiles of visitors to
interesting places and 610 hotel guests in Mers-
ing in Johor Bahru. As mentioned before, our
54 International Journal of Asian Business and Information Management, 5(4), 48-59, October-December 2014
case study is located in Johor Bahru. Johor Bahru
is the second-largest city in Malaysia, located
in the southern part of the country. The city is
very close to Singapore; thus, Singaporeans can
reach it quite easily. Tourism has become one
of the most significant key areas of economic
growth for Johor Bahru. There are several places
of interest in Johor Bahru including: Legoland,
Danga Bay, Johor Zoo, etc. Also, the Mersing
area is located in Johor Bahru; this is an area
which has a considerable amount of hotels and
resorts for tourists. Thus, this research intends
to conduct a study of the Johor Bahru tourism
industry based on interesting places and Mers-
ing hotels. Through analyzing tourist data of
Johor Bahru, the results deduced varied and
useful information.
4.1. Analysis of Results
The process of analysing tourist data com-
menced by collecting data from the www.tri-
padvisor.com site by means of the RK crawler.
RK crawler is able to obtain the place name and
search the tripadvisor database for this name. It
then crawls visitor data for information such as,
specifically: age, gender, country, comments,
etc. According to the number of visitors, the
length of time required for gathering the data will
be rendered variable by the crawler. In addition,
RK crawler can load the extracted data in CSV
format automatically. Collected data needs to be
clean and normalized in order to be analyzed by
data mining software. The format of gathered
data is then changed to ARFF, which is compat-
ible with Weka data mining software. Before
mining the data, some statistical information can
be extracted, for example: frequency of visitors
based on country of origin or dispersion of them
based on age and other attributes. However, the
hidden rules between their attributes are not
always obvious until completion of data mining;
the data; thus, data mining is the next step to
finding the hidden rules between attributes in
dataset through clustering and association rules
techniques. Likewise, qualitative analysis was
fulfilled by Nvivo software by means of coding
the tourist’s comments. The model illustrates the
Copyright © 2014, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
relationship between impact factors of places
and tourists’ preferences.
4.2. Statistical Standpoint
of Tourists Information
In this part, the tourist data will be surveyed
from the statistical information aspect. Figure
2 illustrates the frequency of visitors by their
country of origin.
According to Figure 2, the largest number of
visitors can be seen as coming from Singapore.
Beside Singaporeans, most of the visitors are
from Malaysia. These numbers are especially
so for visitors to Mersing hotels. Mersing ho-
tels are almost empty of Middle Easterners but
full of Singaporeans, due to the short distance
between the two places. Figure 3 displays the
factor based on age groups. Mersing Island is a
place that potentially attracts visitors because of
its natural beauties. Our result shows that most
visitors who visit this island are from Singapore,
Malaysia, and the UK. There are high levels of
interest in different activities; results show that,
apart from visiting beaches and historical places,
having good food and wine, as well as having a
choice of outdoor and adventure activities are
very attractive options for visitors. Our find-
ings also show that there are some differences
between types of visitors. For instance, males
have different hobbies compared with females.
In addition, some “hidden facts” revealed that if
a visitor is interested in one particular activity,
he/she may have interest in some other similar
activities. In the following sections, we will
use different kinds of data mining techniques
to uncover these interesting rules.
Figure 3 shows the overall rate of reviews.
In addition, it indicates three factors, namely:
cleanliness, service and value grouped by
gender, as well as age group. The figure distin-
guishes satisfaction of different groups based on
these factors. Results show that persons in the
age group of 18 to 24 display less satisfaction
than all other groups according to the average
review rates; while the age groups of 35 to 49,
25 to 34, and 65+ respectively tend to have
greater levels of satisfaction. We discovered a
 International Journal of Asian Business and Information Management, 5(4), 48-59, October-December 2014 55

Figure 2. Frequency of visitors sorted by their country of origin

Figure 3. Factors by age and gender group

Copyright © 2014, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
56 International Journal of Asian Business and Information Management, 5(4), 48-59, October-December 2014

further fact about the 65+ age group, namely:
that those visitors in this age group are not as
satisfied with the six factors, as were the other
age groups.
It was found that Malaysian and Singa-
porean tourists are more interested in doing
shopping during their travel. Even though there
are many tourists from United Kingdom, the
interest level is still very low. Another notable
fact identified is that Belgians are very inter-
ested in casinos and amusement parks. A further
interesting result for us during this analysis
was that Chinese visitors to Mersing hotels
are exclusively female; while French, German,
Indonesian and Dutch tourists are mostly male.
Surprisingly, Dutch tourists were not found to
be interested in outdoor and adventure activi-
ties. Similarly, Malaysians and Singaporeans
are more interested in shopping and spas.
Within the hotel visitors’ profiles, there were
12 activity interests mentioned. In addition, the
results show that a high percentage of visitors
are interested in particular activities such as
visiting the beach, outdoor activities, having
great food etc. Hence, we did not use them in
the analysis because it was obvious to us that
they are favorite activities. However, some
activities with a low interest level may still have
some hidden relationships with each other. This
kind of information was revealed after statistical
analysis was carried out on the dataset; however,
for the purposes of acquiring the extra interest-
ing knowledge from our dataset, we needed to
identify the important groups of tourists based
on their attributes. It was considered that the
clustering technique would be helpful in this
research to detect the most important clusters.
In the following section, clustering techniques
will be applied by Weka in order to recognize
notable clusters of tourists.
4.3. Exploring Notable
Clusters of Tourists
It was recognized that detection of significant
clusters of tourist profile data would be helpful
for us to gain more knowledge in the realm of
tourism. We executed the K-Means and EM
algorithms in order to determine the existence
of some clusters from visitor data. After clus-
tering our dataset, we discovered that some
clusters were especially important in gaining
more knowledge.
As a result of the clustering section, we
found that some clusters can be helpful for
us in this research in terms of recommending
especial places for each of the clusters. Table 1
illustrates a summary list of findings from the
clustering section.
4.4. Interesting Rules
Relating to Tourist Data
By escalating the number of attribute findings,
identifying the rules between them becomes an
increasingly difficult task. However, some rules

Table 1. A list of findings from clustering techniques

Cluster Finding
Singaporean females aged between 25 to 34 and who travel for fun are usually interested in spas,
1 shopping, good food and wine, outdoor activities, historical and cultural places as well as the beach
and sun.
Malaysian females aged between 35 to 49 and who travel for work/fun tend to be interested in spas,
2
shopping, good food and wine as well as the beach.
Singaporean male visitors aged between 35 to 49 and who travel for fun, are interested in good food
3 and wine, outdoor and adventure activities, visiting museums, historical and cultural places, as well as
the beach and sun.
British female visitors in the age group of 25 to 34 who travel for fun tend to be interested in good
4
food and wine, visiting museums as well as historical and cultural places and the beach.

Copyright © 2014, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
 International Journal of Asian Business and Information Management, 5(4), 48-59, October-December 2014 57

are not intuitive in a tourist’s dataset. In this
instance, association rules mining techniques
can lead to the appearance of some hidden rules
among items in the dataset. An Apriori technique
is then applied to uncover hidden relations
between elements in the dataset. Amongst nu-
merous rules, we were able to discover several
useful and valuable ones, some of which are
listed below in Table 2.
As can be seen from the above rules, Aus-
tralian female tourists enjoy visiting Legoland;
while Singaporean male tourists tend to prefer to
have an option of good food and wine, as well
as outdoor activities. These kinds of findings
were not detected through clustering techniques;
therefore, association rules lead us to gain more
interesting knowledge from tourists’ respective
datasets. In the next step, we proceed to analyze
the same dataset using qualitative analysis
by Nvivo. Through the application of Nvivo,
the comments of tourists will be analyzed to
catch the salient points of comments made. We
hope to support both findings using different
approaches.
4.5. Textual Analysis of
Tourist’s Profile
The Nvivo qualitative analysis was carried out
to in order analyze the comments tourists tend
to make about various places. Using Nvivo
actually helped us to make our qualitative
research more transparent; it provided us with
the ability to better address some of the key
challenges in professional communication re-
search, including: efficiency, multiplicity and
transparency (Hoover, 2011; Denardo, 2002).
We coded tourists’ comments so as to extract
a model based on some attributes. The figure
below is the output model of Nvivo achieved
after coding tourists’ comments based on the
impact factors of Legoland and the country of
origin of visitors.
Affirmative points argue the need for
a suitable place to be enjoyed by the whole
family, including children; while the negative
points present liabilities of Legoland, such as
high food prices and hot weather. A number of
results can be extracted from this model. For
example, Australian tourists concentrate on
positive points (such as its being a fun place
for children and great for the family); whereas
Indonesians argue that the prices for food are
expensive. An interesting point gleaned from
this figure is that, in spite of there being negative
points, some foreign visitors (notably Singa-
poreans and Australians) only considered the
positive points. Likewise, word frequency can
be achieved through Nvivo for identifying the
most frequently-used words and their synonyms
in tourists’ comments. In this case, the results
indicate that a great bulk of visitors emphasize
positive points of Legoland; which was the same
result as in the model. Nvivo analysis supports
our result in the Weka analysis section; therefore,
these two analysis techniques can be useful in
the tourism industry in order to acquire definite
results of tourist data.
5. RECOMMENDATIONS
FOR TOURISTS
According to the results provided in previous
sections, we can recommend a list of appropri-
ate places and services for specific groups of

Table 2. List of findings from association rules techniques

Rule Finding
If a tourist is from Australia, female, and aged 35 to 49, then she will most likely select Legoland as a fun
1
travel destination.
2 If tourists are aged 25 to 34 and choose Sanrio Hello Kitty Town, they are most probably females.
If a hotel visitor is from Singapore, male, and interested in good food and wine, he is also likely to be
3
interested in outdoor and adventure activities.

Copyright © 2014, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
58 International Journal of Asian Business and Information Management, 5(4), 48-59, October-December 2014
tourists based on their preferences. For places
of interest, we can suggest the model below
for visitors. This model demonstrates suitable
places for each group of tourists; for instance,
Legoland is one of the best options for Ma-
laysian, Singaporean and Australian travelers.
According to the findings in the associa-
tion rules section in Table 2, we conclude that
Legoland can be included on our recommen-
dation list for Australian female tourists, as
well as Malaysian and Singaporean adults.
It is proposed that this model would be very
beneficial if used by hotel managers. The first
rule displays that, since the spa area is of great
interest for Asians, as well as female visitors,
it must then be recommended to this group of
visitors. The second rule illustrates that pack-
ages of good food and wine can be offered to
Singaporean and spa visitors.
6. CONCLUSION AND
FUTURE WORK
This study has applied quantitative and qualita-
tive analysis together in an attempt to discover
hidden knowledge from tourist profiles in www.
tripadvisor.com. Several data mining and tex-
tual analysis techniques (such as visualization,
clustering, association rules mining and Nvivo
qualitative analysis) have been carried out to
show the performance of data mining in the
tourism industry. In essence, we have shown that
Nvivo analysis in part can support the findings
Copyright © 2014, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
of the Weka analysis section of our research
so as to gain conclusive results. Findings of
this study can be used by tourist associations
and hotel managers in Johor Bahru in order to,
specifically: identify tourist behavior patterns,
improve their services, offer specific travel
packages and, finally, enhance the number of
tourists arriving in Malaysia.
In the future, we would like to carry out
further experiments, in particular, enlarging the
scope of study for the mining of voluminous
amounts of data. Accordingly, it would be very
helpful to propose an automated process by
which to collect and prepare data with suitable
format autonomously. In this respect, the authors
would not have to pre-process and convert the
format manually. Crawled data would actually
be ready to be analyzed through just a few clicks
in a short space of time. Hence, it would provide
an opportunity to analyze the tourist profile of
all interesting spots and hotels in the world in
a short time-frame and easy manner.
ACKNOWLEDGMENT
The author would like to express gratitude for
the thoughtful suggestions and warm support
received from the postgraduate students in the
field of IT Management at Universiti Teknologi
Malaysia (UTM). He would also like to thank
his dear parents for their love and encourage-
ment. UTM is gratefully acknowledged for its
Research University Grant (vot 07J79).
 International Journal of Asian Business and Information Management, 5(4), 48-59, October-December 2014 59

REFERENCES Isa, D., Kallimani, V. P., & Lee, L. H. (2009). Using

Agrawal, R. (1996). Fast discovery of association
rules. Advances in knowledge discovery and data
mining (Vol. 12, pp. 307-328).
Au, N., & Law, R. (2002). Categorical classification
of tourism dining. Journal of Annals of Tourism
Research, 29(3), 819–833. doi:10.1016/S0160-
7383(01)00078-0
Bazeley, P. (2013). Qualitative data analysis with
Nvivo. Sage Publications Limited.
Bose, Data mining in Tourism, 2009, Elsevier.
Choi, T. Y., & Chu, R. (2000). Levels of satisfaction
among Asian and Western travellers. International
Journal of Quality & Reliability Management, 17(2),
116–132. doi:10.1108/02656710010304537
Denardo, A. M. (2002). Using NVivo to analyze
qualitative data.
Etemadi, R. (2010). An approach in web content
mining for clustering web pages (pp. 279–284).
IEEE. doi:10.1109/ICDIM.2010.5664660
Guoxia, Z. (2009). The application of data mining
in tourism information. IEEE, 3, 689-692.
Han, J., & Kamber, M. (2006). Data mining: concepts
and techniques. Elsevier.
Hokey, M. (2002). A data mining approach to devel-
oping the profiles of hotel customers. International
Journal of Contemporary Hospitality Management,
14(6), 274–285. doi:10.1108/09596110210436814
the self organizing map for clustering of text docu-
ments, Journal. Expert Systems with Applications,
36(5), 9584–9591. doi:10.1016/j.eswa.2008.07.082
Kosala, R., & Blockeel, H. (2000). Web mining re-
search: A survey. Journal ACM Sigkdd Explorations
Newsletter, 2(1), 1–15. doi:10.1145/360402.360406
Larose, D. T. (2005). Discovering knowledge in
data: An introduction to data mining. Available at
www.wiley.com
Li, J. (2012). Exploring the Destination Image of
China through International Urban Tourism. Uni-
versity of Waterloo.
Poon, W.-C., & Lock-Teng Low, K. (2005). Are trav-
ellers satisfied with Malaysian hotels? International
Journal of Contemporary Hospitality Management,
17(3), 217–227. doi:10.1108/09596110510591909
Richards, L. (1999). Using NVivo in qualitative
research. Sage (Atlanta, Ga.).
Romero, C., Ventura, S., Zafra, A., & Bra, P.
(2009). Romero, Cristóbal VenturaN, Applying
Web usage mining for personalizing hyperlinks in
Web-based adaptive educational systems. Comput-
ers & Education, 53(3), 828–840. doi:10.1016/j.
compedu.2009.05.003
Witten, I. H., & Frank, E. (2011). Data mining Practi-
cal machine Learning Tools and Techniques. Elsevier.
Zhou, Y. (2008). Constructing tourism association
words set based on association rule mining. IEEE,
1, 571-574.

Hoover, R. S. (2011). Using NVivo to Answer the

Challenges of Qualitative Research in Professional
Communication: Benefits and Best Practices Tutorial,
Professional Communication. IEEE Transactions
on, 54(1), 68–82.

Copyright © 2014, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.