Expert Systems with Applications 36 (2009) 11760–11763

Contents lists available at ScienceDirect

Expert Systems with Applications

journal homepage: www.elsevier.com/locate/eswa

Predicting tourism loyalty using an integrated Bayesian network mechanism

Chi-I Hsu a, Meng-Long Shih b, Biing-Wen Huang c,*, Bing-Yi Lin a, Chun-Nan Lin d
a
Dept. of Information Management, Kainan University, No. 1 Kainan Road, Luchu, Taoyuan County 338, Taiwan
b
Dept. of Social Studies Education, National Taitung University, Taiwan
c
Dept. of Applied Economic, Chung-Hsing University, Taiwan
d
Dept. of Tropical Agriculture and International Cooperation, National Ping-Tung University of Science and Technology, Taiwan

a r t i c l e i n f o a b s t r a c t

Keywords: For effective Bayesian networks (BN) prediction with prior knowledge, this study proposes an integrated
Tourism management BN mechanism that adopts linear structural relation model (LISREL) to examine the belief or causal rela-
Loyalty tionships which are subsequently used as the BN network structure for predicting tourism loyalty. Four
Bayesian networks hundred and fifty-two valid samples were collected from tourists with the tour experience of the Toyugi
Linear structural relation model
hot spring resort, Taiwan. The proposed mechanism is compared with back-propagation neural networks
(BPN) or classification and regression trees (CART) for 10-fold cross-validation. The results indicate that
our approach is able to produce effective prediction outcomes.
Ó 2009 Elsevier Ltd. All rights reserved.

1. Introduction ecology and experience, and the information of agricultural tour-

ism. It is a recreational village with different features.
With the prevalence of tourism in Taiwan, the development of The data analysis was conducted in two stages: verifying rela-
tourism industry has facilitated local economy and increased the tionships in the research model and predicting the level of tourism
employment opportunity. Thus, tourism becomes an industry that loyalty. In the first stage, LISREL was used to verify the belief or
is valued and actively developed by the government. However, causal relationships in the research model of tourism loyalty. LIS-
when facing a more competitive tourism environment, how to at- REL is a structural equation modeling (SEM) technique used to
tract the customers and further transform them into loyal ones will determine whether a research model is valid by examining the
be the key for the operation of leisure business. This research seeks goodness-of-fit between the model and raw data. It has been
to determine the factors that influence tourism loyalty. Moreover, widely applied in social-science research. In the second stage, the
this study proposes an integrated mechanism that combines LIS- supported relationships of the LISREL analysis are used as the BN
REL (Joreskog & Sorbom, 1993) and BN (Pearl, 1986) to predict a network structure to predict a tourist’s loyalty level. The predicted
tourist’s loyalty level. results were also compared with those generated by BPN and
Valid samples were collected from 452 tourists with the tour CART.
experience of the Toyugi Hot Spring Recreational Village, which
is located at the eastern region of Taiwan. The village is managed 2. Tourism loyalty
by Taitung County Farmers Association. With an area of 15 hectare,
it is the largest hot spring recreational village in Chihben hot spring Loyalty refers to the repurchase will of certain products and ser-
area. The village is full of rich ecological resources such as spatial vices (Jones & Sasser, 1997). Fornell, Johnson, Anderson, Cha, and
grasslands, varied plants, wild birds and butterflies. Therefore, Bryant (1996) suggested that ‘‘repurchase possibility” and ‘‘repeti-
the village provides the visitors the combined leisure functions tive purchase” are two critical factors for loyalty assessment. Sir-
such as recreation, conference, experience, education and training ohi, Mclaughlin, and Wittink (1998) indicated the following as
by the unique hot spring, landscapes and farm produces. Visitors the indexes to assess loyalty: (1) continuous purchase; (2) increase
can enjoy the services, herbs and various agricultural products in purchase in the future; (3) recommendation for others’ purchase.
hot spring hotels. Besides, the village also provides junior high This research defines tourism loyalty as the visitors’ will to revisit
schools and elementary schools a teaching space for rural village and recommend the destinations to others after arriving at the
tourist attraction. The measurement indexes include continuous
revisiting, revisiting will and recommendation to others.
* Corresponding author. Tel.: +886 422840349; fax: +886 422860255.
Loyalty is one of the targets of strategic marketing and it allows
E-mail addresses: [email protected] (Chi-I Hsu), [email protected]
(M.-L. Shih), [email protected] (B.-W. Huang), m93221001@ companies to enhance the competitive advantages (Craft, 1999).
ms2.knu.edu.tw (B.-Y. Lin), [email protected] (C.-N. Lin). The benefits of loyalty include below: (1) customers’ repurchase

0957-4174/$ - see front matter Ó 2009 Elsevier Ltd. All rights reserved.
doi:10.1016/j.eswa.2009.04.010
 C.-I Hsu et al. / Expert Systems with Applications 36 (2009) 11760–11763 11761

and promotion willingness can lead to revenue growth of the firms to verify the hypothesized relationships, but it is seldom combined
and the increase of market share; (2) reduction of costs; (3) in- with other machine-learning algorithms. This study uses LISREL to
crease of employees’ work satisfaction (Jacoby, 1994). In order to aid BN in discovering a suitable network architecture for
increase the customers’ loyalty, companies need positive customer prediction.
relationship management (CRM). CRM means the enterprises find
the customers’ real needs with the support of process and technol- 3.1. LISREL
ogy, and improve the products and services that are devoted to the
enhancement of customer loyalty (Kalakota & Robinson, 1999). LISREL is one SEM technique which combines the concepts of
Spengler (1999) also suggested that CRM integrates planning, mar- both factor analysis and path analysis. It is especially appropriate
keting and customer service by information technology, and pro- to use LISREL to analyze the data in social and behavioral research
vides customized services to increase customer loyalty and fields. While multiple regression can estimate the parameters of
corporate operational benefits. In addition, Hui, Wan, and Ho only one linear equation at a time, LISREL can simultaneously pro-
(2007) indicated the characteristics of tourist attractions, such as cess multiple sets of variable relationships to estimate the param-
interesting cultures, attractive urban sightseeing, interesting night eters in an entire system of linear equations in a model. The LISREL
life and attractive natural and scenic aspects might increase cus- model and equations are shown in Fig. 1 in which E is disturbance;
tomer satisfaction and revisiting will. According to the literature g is the vector of endogenous latent variable; a is intercept; B is the
review discussed above, this research proposes three factors which matrix of regression coefficient for endogenous latent variable; C is
might increase tourism loyalty: customer service, web function the matrix of regression coefficient for exogenous latent variable; n
with the support of technology and local characteristics. is the vector of exogenous latent variable; and f is the vector of
Three factors influence travelers’ tourism loyalty are proposed latent disturbance. The analysis consists of two steps: (1) measure-
based on a literature review, including (1) Customer Service (CS): ment model analysis, which aims to analyze the loading relation-
the service consumers received from employees; (2) Web Function ships between latent variables and their corresponding
(WF): the functions providing by the tour web site; (3) Local Char- observable variables, and (2) structural model analysis, which aims
acteristics (LC): the consumer’s perception of the local tourism to analyze the hypotheses relationships among latent variables.
characteristics; and Tourism Loyalty (TL): the loyal degree regard-
ing a tourist revisit a destination. This study suggests that the 3.2. BN
greater the degree to which a tourist perceived regarding customer
service, web function and local characteristics of the destination, A BN is a graphical model of variables and their relationships
the greater is his/her tourism loyalty, which refers to repeat of vis- based on probability theory. It is also called a belief network or
it, willing to revisit and recommend to others. Therefore, H1–H3 is causal network. A BN uses prior probabilities and probabilities in
established as below: sample space to estimate posterior probabilities. In a BN graph, ar-
rows between nodes are used to represent a directed acyclic graph
H1: Customer service positively influences tourism loyalty. (DAG) (Niedermayer, 1998). Each parent node represents the cause
H2: Web function positively influences tourism loyalty. of an event, a child node represents the outcome, and an arrow
H3: Local characteristics positively influences tourism loyalty. represents the causal relation. As shown in Fig. 2, the set of parent
nodes of a node TL is denoted by parents(TL) and the joint distribu-
All questionnaire items are shown in Table 1. Each tourist was tion of the node values can be written as the product of the local
asked to rate on a scale of 1–5 his or her degree of agreement with distributions of each node and its parents.
each item. Advantages of BN include the ability to analyze problems with
incomplete data and to combine domain knowledge and data
3. The integrated Bayesian network mechanism (Hackerman & Wellman, 1995). However, without prior under-
standing or knowledge about the problem domain, the required
To overcome the difficulty of constructing a BN structure when significant computational effort of an NP-hard task in exploring a
learning from data, this study proposes a novel approach that com-
bines LISREL and BN to predict a tourist’s loyalty level. LISREL is an
advanced statistical technique in the social and behavioral sciences E1 CS1
V1

E2 CS2
V2
Table 1 CS
E3 V1
CS3
The questionnaire items.

Factor Item Content E4 CS4

Customer service CS1 Quick response to customers’ suggestions E5 CS5

CS2 Fluent service flow TL1 E12
CS3 Understanding customers’ need E6 WF1
CS4 Quickly solving customers’ problems TL TL2 E13
CS5 Listening to customers’ complaint WF
E7 WF2
TL3 E14
Web function WF1 Web information attracted people
WF2 Ease of use E8 WF3
WF3 Web site security
E9 LC1
Local LC1 Farm products have local characteristics
characteristics LC2 The brand and image of farmers’ association is
relieved E10 LC2 LC
LC3 Service and produce have geographic features
E11 LC3
Tourism loyalty TL1 Repeat of visit
TL2 Revisiting will
η = αη + Bη + Γξ + ς
TL3 Recommendation to others
Fig. 1. LISREL: verifying the believe relationships.
11762 C.-I Hsu et al. / Expert Systems with Applications 36 (2009) 11760–11763

collected after deducting the sample with more than three ques-
CS WF LC tions not answered. The effective response rate was 82.2%.
The resultant sample was then randomly split into two sub-
samples. The first subset was used for exploratory factor analysis
(EFA) to identify the factor structure hidden in the data collected.
TL The other subset was used for LISREL analysis. SPSS 13.0 for Win-
dow version was used for EFA analysis. As shown in Table 2, using
the Eigen value rule, a 4-factor structure emerged.
P (CS ,WF , LC , TL ) = ∏ P (TL parents (TL)) The reliability and validity are examined as shown in Table 3.
The Cronbach’s a values for all constructs are all above the 0.7 level
Fig. 2. BN: predicting the outcome level. (Cronbach, 1947). We examined convergent validity with the com-
posite reliability (CR) and the average variance extracted (AVE) by
the constructs. A CR greater than .60 is preferred (Fornell & Larcker,
Table 2
The result of EFA.
1981) and all constructs in this study met this requirement. The
AVE for all constructs in this study exceeded the preferred 0.5 (For-
Factor Item 1 2 3 4
nell & Larcker, 1981).
Customer service CS1 .796 .128 .014 .181
CS2 .692 .284 .061 .065 4.2. Verifying relationships in the research model
CS3 .648 .380 .301 .123
CS4 .751 .081 .281 .159
CS5 .700 .188 .364 .141 We employed LISREL to examine the measurement and struc-
Web function WF1 .274 .794 .106 .251
tural models. Regarding whether the measurement model and
WF2 .191 .895 .155 .142 structural model are good of fit, the criteria of goodness-of-fit mea-
WF3 .258 .742 .259 .220 sures are as follows: v2 =df is suggested to be smaller than 2 for
Local characteristics LC1 .131 .205 .764 .257 good of fit (Carmines & McIver, 1981) and 3 for acceptable fit (Chin
LC2 .287 .108 .649 .262 & Todd, 1995); CFI is suggested to be greater than 0.95 for good fit
LC3 .158 .142 .766 .113 (Bentler, 1995); NFI, NNFI greater than 0.9 for good fit (Hu & Jen,
Tourism loyalty TL1 .044 .256 .240 .701 2005); IFI greater than 0.9 for good fit (Hu & Jen, 2005); GFI and
TL2 .163 .191 .145 .821 AGFI greater than 0.9 for good of fit, and 0.8 for acceptable fit (Sub-
TL3 .265 .100 .211 .769
hash, 1996); PGFI greater than 0.5 for good fit (Mulaik, James, Van
Altine, Bennett, & Stilwell, 1989); SRMR smaller than 0.08 for good
fit (Hu & Jen, 2005); and RMSEA smaller than 0.05 for good fit and
Table 3
Cronbach’s a, CR, and AVE. 0.08 for acceptable fit (McDonald & Ho, 2002). The goodness-of-fit
indexes for both measurement and structural models are accept-
a CR AVE
able as shown in Table 4.
CS 0.849 0.899 0.642 As shown in Table 5, all path coefficients are significant at the
WF 0.873 0.785 0.549
0.05 level in the structural model. The results indicate that the
LC 0.731 0.764 0.520
TL 0.770 0.875 0.700 hypothesized relationships are supported. Square multiple correla-
tions (SMC) are also reported, which is 0.73 indicating the
explained proportion of variance in tourism loyalty.
previously unknown network is costly and inefficient (Niederma-
yer, 1998). 4.3. Predicting the level of tourism loyalty
BN is becoming an increasingly important solution for practical
problems in the field of Artificial Intelligence (Korb & Nicholson, In the above section, we describe LISREL analysis employment
2003). The applications of BN include the areas of maintenance to verify the hypothesized relationships in the proposed research
project delays based on specialists experience (de Melo & Sanchez, model. Subsequently, this study adopts the supported relation-
2008), detecting firms that issue fraudulent financial statements ships as the BN network structure to predict a tourist’s loyalty le-
(FFS) and identifying factors associated to FFS (Kirkos, Spathis, & vel. Based on the LISREL analysis results, the nodes CS, WF, and LC
Manolopoulos, 2007), and other problem domains. represent the antecedents of the outcome node TL. The input data
for each node is the average value of the corresponding question-
4. The experiment results naire items. The BN software Netica 2.05 (NORSYS, 2000) was used
to construct the probability model using the 452 valid samples. The
4.1. Sample and exploratory factor analysis same software was used to predict the outcome. All the learning/
testing results were obtained for 10-fold cross-validation.
This study develops a questionnaire based on the proposed For comparison purposes, the BPN and CART modules in SPSS-
items and delivered to 550 travelers with the tourism experience Clementine 8.1 software package were used to predict the same
of the Toyugi hot spring resort of farmers’ association, Taitung outcome TL. The system’s default control parameters were
County, Taiwan. Four hundred and ninety-eight travelers replied adopted. The input variables are the same constructs in the re-
the questionnaires. Four hundred and fifty-two valid samples were search model including CS, WF, and LC.

Table 4
The goodness-of-fit indexes.

v2 =df CFI NFI NNFI GFI AGFI PGFI RMSEA SRMR

Criteria <3 >0.95 >0.90 >0.90 >0.80 >0.80 >0.50 <0.08 <0.08
Measurement model 1.80817 0.99 0.97 0.98 0.99 0.89 0.63 0.060 0.042
Structural model 1.80817 0.99 0.97 0.98 0.92 0.89 0.63 0.060 0.042
 C.-I Hsu et al. / Expert Systems with Applications 36 (2009) 11760–11763 11763

Table 5 is relevant to real-world problems. Our approach is able to explain

The hypothesis tests summary. the correlative relationships between variables and outperforms
Hypothesis Content Path coefficient t-Value Support BPN and CART by achieving higher predictive accuracy. BPN and
H1 CS ? TL 0.30 2.80 Yes CART are appropriate for exploratory studies in which the relation-
H2 WF ? TL 0.18 2.06 Yes ship between variables is unclear, whereas our approach is suitable
H3 LC ? TL 0.46 3.60 Yes for data prediction in empirical research with a theoretical basis.
Because constructing a BN structure when learning from data
presents certain difficulties (Niedermayer, 1998), the approach
proposed in this research is demonstrated to be a prospective
Table 6
The experiment results.
way to aid a BN in discovering a suitable network architecture with
better prediction performance. In the future, a sensitivity analysis
LISREL–BN BPN CART would be helpful to understand the approach robustness. Further
Training Testing Training Testing Training Testing simulations are needed to be of general interest. As for the contri-
1 0.143 0.239 0.667 0.804 0.667 0.761 bution of this research, LISREL is a widely used advanced statistical
2 0.265 0.348 0.663 0.733 0.673 0.489 tool in the social and behavioral sciences, but it is seldom com-
3 0.263 0.370 0.688 0.511 0.715 0.400 bined with other machine-learning algorithms. The proposed ap-
4 0.280 0.217 0.663 0.444 0.678 0.178
proach can be used as a good reference for related research in
5 0.285 0.174 0.690 0.600 0.690 0.533
6 0.278 0.239 0.666 0.733 0.676 0.444 social-science fields.
7 0.265 0.348 0.666 0.756 0.676 0.556
8 0.263 0.370 0.690 0.622 0.715 0.400 References
9 0.283 0.196 0.649 0.289 0.681 0.178
10 0.283 0.196 0.690 0.609 0.687 0.413 Bentler, P. M. (1995). EQS structural equations program manual Multivariate Software.
MAE 0.261 0.270 0.673 0.610 0.686 0.435 CA: Encino.
Carmines, E. G., & McIver, J. P. (1981). Analyzing models with unobserved variables:
Analysis of covariance structures. In G. W. Bohrnstedt & E. F. Borgatta (Eds.),
Social measurement: current issues. Beverly Hills, CA: Sage.
Chin, W. W., & Todd, P. (1995). On the use, usefulness and ease of use of structural
The 10-fold cross-validation results for the mean average errors equation modeling in MIS research: A note of caution. MIS Quarterly, 19(2),
(MAE) for LISREL–BN, BPN and CART are shown in Table 6. It is 237–246.
indicated that our approach obviously outperforms BPN and CART Craft, S. H. (1999). Marketers gain by measuring true loyalty. Marketing News, 33, 18.
Cronbach, L. (1947). Test ‘reliability’: Its meaning and determination. Psychometrika,
in both the training (.261 vs. .673 and .686) and testing results 16, 1–16.
(.270 vs. .610 and .435) with 10-fold average basis. de Melo, Ana C. V., & Sanchez, Adilson J. (2008). Software maintenance project
delays prediction using Bayesian Networks. Expert Systems with Applications,
34(2), 908–919.
5. Conclusions Fornell, C., & Larcker, D. F. (1981). Evaluating structural equation models with
unobservables and measurement error. Journal of Marketing Research, 18(1),
5.1. Tourism implications 39–50.
Fornell, C., Johnson, M. D., Anderson, E. W., Cha, J., & Bryant, B. E. (1996). The
American customer satisfaction index: Nature, purpose, and finding. Journal of
According to the LISREL results, we found that: (1) customer ser- Marketing, 60(spring), 7–18.
vice, web function and local characteristics have significant and po- Hackerman, D., & Wellman, M. P. (1995). Bayesian networks. Communications of the
ACM, 38(3), 27–30.
sitive influence on tourism loyalty. Local characteristics are the Hu, K. C., & Jen, W. (2005). Applications of LISREL and neural network to analyze the
most important, followed by customer service and web function. passenger’s behavioral intentions. Logistics Research Review, 8, 43–55.
In order to increase tourism loyalty to attract visitors’ revisiting Hui, T. K., Wan, D., & Ho, A. (2007). Tourists’ satisfaction, recommendation and
revisiting Singapore. Tourism Management, 28, 965–975.
or recommendation to others, and increase the number of visitors,
Jacoby, R. (1994). Why some customers are more equal than others. Fortune, 130,
the key of tourism management is the reinforcement or use of the 9–13.
characteristics of the spots. With regard to customer service, tour- Jones, T. O., & Sasser, W. E. Jr., (1997). Why satisfied customers defect. Harvard
Business Review, 73(16), 88–99.
ism management should value the quick response to customers’
Joreskog, K., & Sorbom, D. (1993). LISREL 8: User’s reference guide. Chicago: Scientific
suggestions and quickly solve the customers’ problems. Finally, Software International.
with the prevalence of Internet business application, the web func- Kalakota, R., & Robinson, M. (1999). E-business: Roadmap for success (first ed.). USA:
tion of tourism websites is commonly used by visitors. The tourism Mary T. O’Brien.
Kirkos, E., Spathis, C., & Manolopoulos, Y. (2007). Data mining techniques for the
administrators should recognize that the enhancement and detection of fraudulent financial statements. Expert Systems with Applications,
improvement of web function will increase tourism loyalty. 32(4), 995–1003.
Korb, K. B., & Nicholson, A. E. (2003). Bayesian artificial intelligence. Chapman & Hall/
CRC.
5.2. Methodology discussion McDonald, R. P., & Ho, M. R. (2002). Principles and practice in reporting structural
equation analysis. Psychological Methods, 7, 64–82.
Although LISREL and BN have been widely applied individually Mulaik, S. A., James, L. R., Van Altine, J., Bennett, N., & Stilwell, C. D. (1989).
Evaluation of goodness-of-fit indices for structural equation models.
in many research studies, few of these studies have investigated Psychological Bulletin, 105, 430–445.
combining them for predictive purposes. This research has pro- Niedermayer, D. (1998). An introduction to Bayesian networks and their contemporary
posed an integrated mechanism to predict a tourist’s loyalty level. applications. <http://www.niedermayer.ca> Retrieved 20.06.06.
NORSYS. Netica 2.05. (2000). <http://www.norsys.com> Retrieved 10.12.06.
The prediction performance of this approach has also been com-
Pearl, J. (1986). Fusion, propagation, and structuring in belief networks. Artificial
pared with the predictions of BPN and CART. Intelligence, 29(3), 241–288.
The three methods LISREL, BN, and BPN represent relationships Sirohi, N., Mclaughlin, E. W., & Wittink, D. R. (1998). A model of consumer
perception and store loyalty intentions for supermarket retailer. Journal of
as networks. However, in terms of explanatory power, the BPN’s
Retailing, 74(2), 223–245.
internal learning is processed in a black-box mode in which the Spengler, B. (1999). Eyes on the customer. Computer World, 33, 60–62.
internal weights are difficult to express in an explicit way which Subhash, S. (1996). Applied multivariate techniques. NY: John Wiley and Sons.