Published on

  • Be the first to comment

  • Be the first to like this

No Downloads
Total views
On SlideShare
From Embeds
Number of Embeds
Embeds 0
No embeds

No notes for slide


  1. 1. USING DATA MINING FOR TRAVEL AND TOURISM RESEARCH Hian Chye Koh School of Business, SIM University 461 Clementi Road, Singapore 599491 Email: hckoh@unisim.edu.sg Gabriel Gervais School of Business, SIM University 461 Clementi Road, Singapore 599491 Email: gabrielgervais@unisim.edu.sg ABSTRACT Data mining can be defined as the process of analyzing mostly large data sets to explore and discover previously unknown patterns, trends and relationships to generate information for better decision making. It has been argued that in today’s fast-paced and competitive environment, for travel and tourism businesses to increase their market share and maintain leadership, these businesses would be hard-pressed not to use data mining tools and techniques to develop, manage and market tourism products and services. In recent years, researchers and practitioners have called for the use of data mining to aid destination management and planning and marketing, market segmentation, customer relationship management and churn modeling. Without doubt, data mining can help travel and tourism businesses to survive, gain a competitive advantage and grow. The objective of this paper is to discuss and illustrate data mining and its application in travel and tourism research. Keywords: Data mining; association analysis; predictive modeling; travel and tourism research INTRODUCTION Data mining, as used today, can be considered a relatively recently developed methodology and technology, coming into prominence only in 1994 (Trybula, 1997). It uses techniques from the disciplines of statistics/mathematics, machine learning and artificial intelligence. It aims to identify valid, novel, potentially useful and understandable correlations and patterns in data (Chung & Gray, 1999) by combing through copious data sets to sniff out patterns and relationships that are too subtle or complex for humans to detect (Kreuze, 2001). Since the mid-1990’s, data mining has been used intensively and extensively by financial institutions (e.g., for credit scoring and fraud detection), marketers (e.g., for direct marketing and cross-/up-selling), retailers (e.g., for market segmentation and store layout) and manufacturers (e.g., for quality control and maintenance scheduling), among others. The increasing popularity and application of data mining can be explained by a few important developments in the last decade. 1
  2. 2. Firstly, advances in both computer hardware and software have made many data mining applications more accessible and affordable to businesses now than ever before. These include cheaper and more powerful computers and more user-friendly, comprehensive and advanced data mining software. Secondly, with the data explosion experienced by many organizations collecting increasingly larger amounts of data (e.g., on transactions and customers), organizations have begun to realize that data are not useful for decision making unless they can be transformed into information, which is an important asset. In this respect, data mining provides the means to analyze large databases to generate valuable business information. Thirdly, success stories of data mining applications reported by businesses as well as aggressive marketing by data mining consultants and software vendors have resulted in increasing numbers of organizations jumping onto the data mining bandwagon. As stated by Davis (1999), more companies are now using data mining as the foundation for strategies that help them outsmart competitors, identify new customers and lower costs. The objective of this paper is to discuss and illustrate data mining and its application in travel and tourism research. DATA MINING METHODOLOGY AND TOOLS The methodology of data mining can be broadly divided into three major stages: pre- modeling, modeling and post-modeling (see Figure 1). Pre-modeling Stage The first step in data mining is to identify the business problem. This is critical as one important reason for failure in data mining is the overemphasis on data analysis at the expense of the business problem to be addressed. However, stating a business problem by itself does not automatically suggest a data mining application. For business problems that are appropriate for data mining, it is necessary to translate the business problem into a data mining application. This is the second step in the pre- modeling stage. The third step is to assess the data needed and available for the data mining application. Data can come from internal sources (e.g., existing customer or transaction records) or external sources (e.g., data compiled from marketing surveys). If the needed data are not available, then they will have to be obtained (e.g., by purchasing the data from data providers) or generated (e.g., by running a test marketing campaign). The fourth and last step in the pre-modeling stage is also the most tedious step – preparation of the data for mining. In many instances, the necessary data are available but not in the same database or in the same standard format. Efforts will then have to be made to extract and combine data from different databases or sources and make them consistent. More importantly, existing data may be incomplete or may contain errors. These data problems have to be dealt with too (say, by filling in missing data from other sources and correcting erroneous data). 2
  3. 3. FIGURE 1. DATA MINING METHODOLOGY Pre-modeling 1. Identification of Business Problem/Issue 2. Translation of Business Problem/Issue into Data Mining Application 3. Assessment of Data 4. Preparation of Data for Mining Modeling 1. Identification of Appropriate Tools/Techniques 2. Analysis of Data 3. Assessment of Results 4. Identification of Final Model Post-modeling 1. Deployment of Data Mining Model/Results 2. Tracking of Performance Modeling Stage The modeling stage can be deemed to be the core of data mining. This is the stage where data analysis is performed. Generally, for any data mining application, several data mining tools or techniques can be used (e.g., logistic regression, neural networks or decision trees can be used for churn modeling to determine customer turnover). Thus, the first step in the modeling stage is to identify the appropriate tools or techniques to use. Once the appropriate tools or techniques are identified, the next step is to perform the actual data analysis (or modeling). After the analysis, it is necessary to assess the results (the third step of the modeling stage). This relates to the objective of the data mining application. For example, if the objective is to perform market segmentation, then it is necessary to assess if the clustering results lead to interpretable, useful and actionable market segments. Assessment of results can be statistical in nature too (e.g., evaluating statistical significance). Data mining is an iterative process. Hence, the assessment may lead to a re-selection of the variables or a re-run of the analysis … etc. Finally, if two or more models give acceptable results, then there is a need to identify the final (or best) model. Thus, the different acceptable models can be compared with respect to their accuracy rates and the one that is most accurate can be selected as the final model. 3
  4. 4. Post-modeling Stage The post-modeling stage relates to the actions to be taken after the data analysis is completed. The first step is the deployment of the data mining model or results. What this involves depends on the objective of the data mining application. For example, if the objective is market segmentation, then the clustering results may be used as inputs for decision making (e.g., designing new products for different market segments, reaching different market segments through different advertisement channels, formulating different strategies for different market segments … etc.). The last step is the tracking of performance. This is necessary because of changes in the environment in which an organization operates. Such changes may lead to a deterioration of the performance of data mining models, which may be dated. For instance, the variables and relationships that help predict a target variable may change over time (e.g., response to a mailing campaign) and a data mining model constructed in the past may no longer be useful at present. Hence, tracking is important as deteriorating performance may signal the need to look at the data mining model again and to build an updated model if necessary. Finally, the double-headed arrows connecting the three stages of the data mining methodology in Figure 1 indicate that data mining is an interactive and iterative process. Very often, it is necessary to move back and forth among stages or steps when developing a data mining application. For example, poor modeling results may mean looking at the data again. Hence, data mining is not a strictly sequential process. Data Mining Tools Data mining tools can be broadly classified based on what they can do, namely: (1) description and visualization; (2) association and clustering; and (3) classification and estimation (i.e., predictive modeling). Some authors (e.g., Berry & Linoff, 2000) have classified data mining tools into more detailed categories. Description and visualization can contribute greatly towards understanding a data set (especially a large one) and detecting hidden patterns in the data (especially complicated data containing complex interactions and non-linear relationships). As such, they are frequently performed before modeling is attempted in order to understand and/or detect relationships among variables. Description and visualization also help greatly in the summarization of data and in the presentation and reporting of results. Description refers to the summarization of data to facilitate understanding. An example of description is the profiling of data sets in order to understand their characteristics, similarities and differences. Standard description tools include summary statistics such as measures of central tendency (e.g., mean), measures of dispersion (e.g., standard deviation) and counts (e.g., cross-tabulation). Graphical approaches (e.g., distributions and plots) can also help to describe data and the relationships in data. Visualization can be considered an enhanced graphical approach that allows user input and interaction. An example is a rotating multidimensional plot that permits the user to define multiple dimensions (i.e., multiple variables) in the plot as well as the direction 4
  5. 5. and angle of rotation to facilitate viewing complex relationships. Colors can also enhance visualization. Association and Clustering In association analysis, the objective is to determine which variables/items go together. It is a tool that looks for groupings or patterns among a set of items. For example, market basket analysis refers to a technique that generates probabilistic statements such as: if a tourist visits Hong Kong, there is a 0.65 probability that he/she also visits China. Such statements (or rules) are intuitive and easy to understand. Also, good association is expected to have predictive value. However, many applications of association analysis are only exploratory in nature, with a view to better understand groupings and patterns in the data set. Association rules can be useful for items bundling, discount and promotion decisions, cross-selling … etc. Association analysis can be extended to include more sophisticated applications. For example, time sequence can be incorporated. To increase on-line purchases, an on-line travel agency may try to isolate the sequences of web navigation that are likely or unlikely to lead to on-line purchases. Actions (e.g., web re-design) can then be taken to promote sales-generating sequences/behavior. Clustering is an exploratory technique that attempts to discover natural groupings in data. The objective is to group similar (homogeneous) objects into the same cluster and dissimilar (heterogeneous) objects into different clusters. Clustering is usually used to do market segmentation and to identify the cluster profile of the different segments. Knowing how the market is segmented and the characteristics of the different segments helps in decisions such as an organization’s competitive positioning, the products to market to particular market segments, and the avenues and communications that can be used to reach the targeted segments. Predictive Modeling The most common and important applications in data mining usually involve predictive modeling, which can be further categorized into two major categories. Classification refers to the prediction of a target variable that is qualitative (i.e., categorical) in nature (e.g., predicting response versus non-response to a mailing campaign). Estimation, on the other hand, refers to the prediction of a target variable that is quantitative (i.e., continuous) in nature (e.g., predicting the amount spent). Generally, predictive modeling attempts to predict a target variable on the basis of one or more input variables. For example, multiple regression can be used to predict the amount of tourist expenditure based on age, gender and income. Neural networks can also be used for predictive modeling. They can often model complex relationships in data well. Neural networks are modeled after the human brain, which can be perceived as a highly connected network of neurons. Finally, decision trees can be used for predictive modeling too. They divide observations into mutually exclusive and exhaustive subgroups based on the levels of particular input variables that have the strongest association with the target variable. The end product can be graphically represented by a tree-like structure (called a decision tree), which is a 5
  6. 6. compact explanation of the data. The end product can also be represented by explicit decision rules. Both representations are easy to interpret and use. Decision trees can model complex relationships reasonably well. In practice, it is common to construct all the regression, neural network and decision tree models and then assess the competing models to identify a final (i.e., best) model. DATA MINING AND TRAVEL AND TOURISM RESEARCH Several factors favor and facilitate, if not necessitate, the use of data mining for travel and tourism research. Firstly, the travel and tourism industry is already one of the largest users of information technology (Sheldon, 1997). Advances in information technology affect the services and amenities offered and how they are delivered and promoted. They also affect the organizational structure and the interactions between customers and service providers (Olsen & Connolly, 1999). In addition, potential and actual travellers are increasingly using sophisticated internet and communication technology to find tours that meet their expectations and needs. Conversely, the internet provides travel and tourism businesses a low cost and more targeted approach to seek customers. Secondly, as concluded by Pyo, Uysal & Chang (2002), for travel and tourism businesses to increase their market share and maintain leadership in today’s fast-paced and competitive environment, these businesses would be hard-pressed not to use data mining tools and techniques to develop, manage and market tourism products and services. Destination management and planning, and marketing need to be more precisely targeted and more aggressive. Buhalis (1999) suggested that an integrated knowledge of tourist characteristics, images, attitudes and preferred destination attributes should be used to market destinations more effectively. Magnini, Honeycutt & Hodge (2003) further suggested that hotels can use data mining to create direct mailing campaigns, plan seasonal promotions, plan the timing and placement of advertisement campaigns, create personalized advertisements, define which market segments are growing most rapidly, and determine the number of rooms to reserve for wholesale customers and business travellers. Several other data mining applications for the travel and tourism industry have also been suggested by Pyo, Uysal & Chang (2002). Accordingly, data mining is one of the most important ways to help travel and tourism businesses to survive, gain a competitive advantage and grow. Thirdly, there is a data explosion in the travel and tourism industry. The proliferation of centralized reservation and property-management systems has resulted in large amounts of consumer data for hotel corporations (Magnini, Honeycutt & Hodge, 2003). At the same time, there is also more access to more data. Beirne (2000) reported that Cendant marketers could access a data warehouse holding data gleaned from the loyalty programs of Cendant’s eight hotel brands, Avis Rental Car and Resort Condominiums International. They could also eventually gain access to Cendant’s direct marketing companies such as AutoVantage, Privacy Guard, Shoppers Advantage and Traveller Advantage. A similar initiative has earlier been reported for Hilton Hotels Corporation and Hilton International (Anonymous, 1999). Without doubt, these massive data have to be transformed into information for decision-making and business intelligence before they can be an asset to businesses. 6
  7. 7. The recent travel and tourism literature documents several cases of data mining applications. The range and scope of these applications are rather diverse. Similar to the earlier classification of data mining tools, these applications can be classified into two broad categories: association and clustering; and classification and estimation (i.e., predictive modeling). No application is classified under the category of description and visualization as this is seldom an end in itself. Description and visualization results are frequently used to understand the data and the patterns and relationships among them better so as to facilitate further analysis. Despite this, description and visualization are very important data mining tools. Small (2000), for example, used pie charts and histograms to show the demographic profile of leisure travelers who use travel agents, the travel characteristics of leisure travellers, and the various sources of information used by travellers when planning for a trip. Association and Clustering Dev, Klein & Fisher (1996) and d’Hauteserre (2000) provided two interesting applications of association analysis. The first relates to the market basket analysis of hotels, airlines, car rental companies and other products and services among tourists for the purpose of partner selection for marketing alliances. The second involves casino resort destinations supplying various amenities and attractions in addition to gaming facilities. In this situation, the market basket analysis of the preferred products among potential visitors is an important analysis to perform before an investment is made. On clustering, Lau, Lee, Lam & Ho (2001) suggested segmenting potential travellers into different clusters based on personal information mined from their personal web sites. This would enable travel and tourism businesses to understand the potential travellers’ interests and needs, and hence be able to offer (via email) specially designed packages. Predictive Modeling Predictive modeling applications in the travel and tourism industry include customer relationship management, cross-/-up selling and churn modeling. For example, data mining can be performed to identify potential tourists who are likely to respond to direct mailing campaigns, guests who under- or over-stay their reservation, or the type of rooms/services that a guest prefers (Kasavana & Knutson, 1999). Also, as reported by Magnini, Honeycutt & Hodge (2003), data mining has been used by Harrah’s hotels and casinos to predict the potential value of each customer, design marketing strategies to retain customers and generate information for customer relationship management. 7
  8. 8. Pyo, Uysal & Chang (2002) discussed a data mining application to identify visitors to Cheju Island (South Korea) who would, upon returning home, recommend to their friends and relatives to visit the island. Stepwise regression was used to regress the intention to recommend (i.e., the target variable) on input variables such as expectations, perceptions, image, experience and demographic variables. The model would be used by the Department of Tourism of the Cheju Provincial Government to distribute appropriately designed promotional material to selected visitors. Finally, Min, Min & Emam (2002) presented a case study involving 11 luxurious hotels in Seoul (South Korea). It detailed the use of decision trees to understand customers’ preferences and interactions with customers in order to build and maintain a loyal customer relationship. ILLUSTRATIONS OF DATA MINING APPLICATIONS From the above, it can be concluded that data mining can contribute greatly to travel and tourism research and to the success of the industry. To illustrate the use of data mining, consider a fictitious company Best Travel Agency (abbreviated as Best from this point to facilitate discussion). Illustration One: Developing Tour Packages Best has observed an increasing number of tourists going to Beijing (China) from the US and wants to capture a greater share of this tourist segment. Currently, the itineraries for Beijing tours cover a long list of the most common and popular tourist attractions. Recent feedback from Best’s customers has indicated that not all these attractions appeal to all customers. Subgroups of customers seem to be interested in only particular subsets of attractions. In addition, many of Best’s customers who are keen to visit other Chinese cities want to have a brief 1- or 2-day stopover in Beijing. To improve its China tour packages, Best wants to find out which tourist attractions in Beijing can be grouped into subsets that would appeal of different subgroups of customers. Best is considering offering a basic China tour package (outside of Beijing) that incorporates different options to additional attractions in Beijing for different subgroups of customers. For this brief stopover in Beijing, only two or three tourist attractions are feasible. After conducting several focus group sessions, the following attractions appear to be the ten most popular places of interest for the Beijing options: (1) Forbidden City; (2) Great Wall of China; (3) Temple of Heaven; (4) Summer Palace; (5) Ming Tombs; (6) Hutong and Courtyard; (7) Cultural Village; (8) Beihai Park; (9) Tiananmen Square; and (10) Beijing Art Museum. To develop the Beijing options further, Best has engaged a marketing research firm in China to conduct a survey of 2000 tourists who have visited Beijing on tour packages of at least five days. As the survey is conducted at the Beijing International Airport, it is kept very simple by merely asking the respondents if they have greatly enjoyed visiting the ten tourist attractions listed above to the extent of wanting to visit them again and/or to recommend them to their relatives and friends. A favourable response 8
  9. 9. for an attraction is taken as an indication of the appropriateness of the attraction for inclusion in the Beijing options. The data collected are then entered into an SPSS file for analysis. SPSS Clementine (a very user-friendly data mining software) is used to do the data mining. The distribution results (see the top panel in Figure 2) show that the five most popular attractions (in descending order of popularity) are: (1) Forbidden City, (2) Summer Palace, (3) Cultural Village, (4) Great Wall of China, and (5) Ming Tombs. Further analysis is performed using the SPSS Clementine Apriori algorithm. The association analysis results are shown in the bottom panel in Figure 2. The association rules can be interpreted as follows. Consider the first association rule listed, which is reproduced below: Instances Support Confidence Lift Consequent Antecedent 1014 50.70 90.500 1.772 Summer Palace Cultural Village The rule can be translated to mean that: (1) Out of the 2000 tourists surveyed, 1014 (or 50.70%) would want to visit Cultural Village again and/or would recommend Cultural Village to their relatives and friends. (2) Out of these 1014 tourists, about 90.50% (or approximately 918 of them) would also want to visit Summer Palace again and/or would recommend Summer Palace to their relatives and friends. (3) Out of the 2000 tourists surveyed, 1022 (or 51.10%) would want to visit Summer Palace again and/or would recommend Summer Palace to their relatives and friends [this can be seen from the second association rule in the panel]. Out of the 1014 tourists mentioned in (2) above, 90.50% would do so. Hence, the lift value is 90.50/51.10 or 1.772. Generally, a higher lift value indicates a more effective association rule. The other association rules can be interpreted similarly. Overall, the association rules suggest the following two Beijing options: (1) Summer Place, Cultural Village and Ming Tombs, and (2) Great Wall of China and Forbidden City. These two options are expected to add the greatest value as a Beijing extension to the regular China tour packages outside of Beijing. Illustration Two: Target Mailing Campaign Best has conducted a mailing campaign a year ago to promote a new tour package to China. The promotion brochure was sent randomly to 2500 customers and 831 customers responded, giving a very good response rate of 33.24%. The data related to this mailing campaign have been entered into a SPSS data set. Best is now considering a target mailing campaign for a similar but new tour package to China. It believes that the patterns and relationships in the data can help it improve the response rate and reduce the cost of the campaign by targeting potential customers better. 9
  10. 10. FIGURE 2. GROUPINGS OF ATTRACTIONS IN BEIJING Note: Reproduced by permission of SPSS Inc. The following demographic characteristics are captured in the database: (1) gender; (2) age group, (3) educational level, and (4) occupation. In addition, the following data are also captured: (5) response to the mailing campaign, (6) whether the customer has taken any tour to China within six months before the mailing campaign, (7) whether the customer has taken any tour to Hong Kong within six months before the mailing campaign, (8) whether the customer has taken any tour within six months before the mailing campaign to countries other than China and Hong Kong, and (9) the number of times the customer has traveled with the agency during the past three years. Given the ease of interpretation of decision tree results, Best has decided to construct a decision tree that can explain/predict customer response. SPSS Clementine is used to construct the decision tree. The results are summarized in Figure 3. 10
  11. 11. FIGURE 3. DECISION TREE RESULTS Note: Reproduced by permission of SPSS Inc. The results show that customers who have responded positively to the last mailing campaign for the China tour package are likely to be: (1) customers who are above 45 years old, female and who have taken a tour within six months before the mailing campaign to countries other than China or Hong Kong [node 10 in Figure 3]; (2) customers who are above 45 years old and who have neither taken a tour within six months before the mailing campaign to countries outside of China or Hong Kong nor to China [node 8]; and (3) customers who are above 45 years old, female and who 11
  12. 12. have taken a tour within six months before the mailing campaign to China but not to countries outside of China or Hong Kong [node 12]. On the other hand, the results show that customers who have responded negatively to the last mailing campaign for the China tour package are likely to be: (1) customers who are aged 45 years and below [node 2]; (2) customers who are above 45 years old and who have taken a tour within six months before the mailing campaign to Hong Kong as well as countries outside of China and Hong Kong [node 5]; (3) customers who are above 45 years old and male and who have taken a tour within six months before the mailing campaign to countries outside of China or Hong Kong but not to Hong Kong [node 9]; and (4) customers who are above 45 years old, male and who have taken a tour within six months before the mailing campaign to China but not to countries outside of China or Hong Kong [node 11]. Except for the variables appearing in the decision tree, the other input variables (i.e., educational level, occupation and number of tours in the past three years) do not appear to be associated with the response to the last mailing campaign. The decision tree results are expected to be useful to Best in its next mailing campaign in targeting the right customers. In this application, it is assumed that Best has an up-to-date database of its customers with respect to the variables captured in the decision tree. CONCLUSION Data mining can be defined as the process of analyzing mostly large data sets to explore and discover previously unknown patterns, trends and relationships to generate information for better decision-making. It is a very powerful and useful methodology and technology. For the travel and tourism industry, data mining can contribute greatly to the ability of travel and tourism businesses to gain a competitive advantage and grow. However, data mining is not without limitations. Some of the major ones are highlighted below. Firstly, the quality of data mining results and applications depends on the availability and quality of data (Chopoorian, Witherell, Khalil &, 2001). Also, the data needed for data mining often exist in different settings and data systems. Hence, they have to be collected and integrated before data mining can be done. In addition, for the data available, problems such as missing data, corrupted data, inconsistent data … etc. have to be resolved before mining is done. It has been estimated that data preparation comprises about 75% of the resources needed for a data mining project. Secondly, a sufficiently exhaustive mining of data may throw up patterns of some kind that are a product of random fluctuations (Hand, 1998). This is especially so for large data sets with many variables. Murray (1997) and Hand (1998) have warned against using data mining for data dredging or fishing (i.e., randomly trawling through data in the hope of identifying patterns). Data mining performed in a mechanical manner will not guarantee results or success. There is a need for human intervention, interpretation and judgment (Pyo, Uysal & Chang, 2002). 12
  13. 13. Thirdly, the successful application of data mining requires the user to be knowledgeable in the domain area of application as well as in the data mining methodology and tools. Domain area knowledge is important as it is critical to identifying the appropriate business issues for which to develop data mining applications (Pyo, Uysal & Chang, 2002). Also, it is essential for specifying the appropriate models and correctly interpreting the results. In addition, without a sufficient knowledge of data mining methodology and tools, the user may not be aware of or be able to avoid the pitfalls of data mining (see, for example, McQueen & Thorley, 1999). IT (i.e., information technology) expertise is also necessary for tasks such as the extraction and preparation of data for mining and the deployment of data mining models (e.g., to embed the models into transactional systems). Further, statistical and research expertise is essential as a data mining project can be considered as a large-scale research project that involves a lot of statistical and research issues. Therefore, collectively, the data mining team should possess the following knowledge and skills: domain knowledge, data mining knowledge and skills, IT expertise, and statistical and research expertise (see also Magnini, Honeycutt & Hodge, 2003). Finally, organizations developing data mining applications need to make a substantial investment of their resources (e.g., time, effort and money) in data mining. Data mining projects can fail for a variety of reasons such as lack of management support and organizational commitment, unrealistic user expectations, poor project management, inadequate data mining expertise … etc. Data mining requires intensive planning and technological preparation work. In addition, top management has to be convinced of the usefulness of data mining and be willing to change work processes, if necessary. Further, all parties involved in the data mining effort have to collaborate and cooperate (Gillespie, 2002). Despite the limitations highlighted above, there is no doubt that data mining can play a critical role in travel and tourism research. What remains is for the travel and tourism industry and businesses to realize the potential benefits and usefulness of data mining. REFERENCES Anonymous. 1999. Hilton, Micros solidify data mining linkage. Lodging Hospitality, 55(13), 88. Beirne, M. 2000. Becoming a true believer. Brandweek, 41(9), 54-56. Berry, M. J. A. & Linoff, G. S. 2000. Mastering Data Mining: The Art and Science of Customer Relationship Management. New York: John Wiley & Sons, Inc. Buhalis, D. 1999. Information technology for small and medium-sized tourism enterprises: adaptation and benefits. Information Technology and Tourism, 2(2), 79-95. Chopoorian, J. A., Witherell, R., Khalil, O. E. M. & Ahmed, M. 2001. Mind your own business by mining your data. SAM Advanced Management Journal, 66(2), 45-51. Chung, H. M. & Gray, P. 1999. Data mining. Journal of Management Information Systems, 16(1), 11-13. Davis, B. 1999. Data mining transformed. InformationWeek, 751, 86-88. 13
  14. 14. Dev, C. S., Klein, S. & Fisher, R. A. 1996. A market-based approach for partner selection in marketing alliances. Journal of Travel Research, 35(1), 11-17. d’Hauteserre, A. M. 2000. Lessons in managed destination competitiveness: the case of Foxwoods Casino Resort. Tourism Management, 21(1), 23-32. Gillespie, G. 2000. There’s gold in them thar’ databases. Health Data Management, 8(11), 40-52. Hand, D. J. 1998. Data mining: statistics and more? The American Statistician, 52(2), 112-118. Kasavana, M. L. & Knutson, B. J. 1999. A primer on software: warehousing, marting, and mining hospitality data for more effective marketing decisions. Journal of Hospitality and Leisure Marketing, 6(1), 83-86. Kreuze, D. 2001. Debugging hospitals. Technology Review, 104(2), 32. Lau, K. N., Lee, K. H., Lam, P. Y. and Ho, Y. 2001. Web-site marketing for the travel-and-tourism industry. Cornell Hotel and Restaurant Administration Quarterly, 42(6), 55-62. Magnini, V. P., Honeycutt Jr., E. D. & Hodge, S. K. 2003. Data mining for hotel firms: use and limitations. Cornell Hotel and Restaurant Administration Quarterly, 44(2), 94-105. McQueen, G. & Thorley, S. 1999. Mining fool’s gold. Financial Analysts Journal, 55(2), 61-72. Min, H. K., Min, H. S. & Emam, A. 2002. A data mining approach to developing the profiles of hotel customers. International Journal of Contemporary Hospitality Management, 14(6), 274-285. Murray, L. R. 1997. Lies, damned lies and more statistics: the neglected issue of multiplicity in accounting research. Accounting and Business Research, 27(3), 243-258. Olsen, M. & Connolly, D. 1999. Antecedent of technological change in the hospitality industry. Tourism Analysis, 4(1), 29-46. Pyo, S., M. Uysal & Chang, H. 2002. Knowledge discovery in databases for tourist destinations. Journal of Travel Research, 40(4), 396-403. Sheldon, P. J. 1997. The Tourism Information Technology. Wallingford: CAB International. Small, S. H. 2000. Fly me to the moon. American Demographics, 22(3), 50-52. Trybula, W. J. 1997. Data mining and knowledge discovery. Annual Review of Information Science and Technology, 32, 197-229. 14