|
Article Excerpt
Abstract
Passenger loads on Taiwanese domestic flights have rapidly declined following the launch of the Taiwan High Speed Rail in 2007. Retaining passenger loyalty is thus a crucial challenge facing Taiwan domestic airlines. This study develops a loyal passenger mining process that is used to assess passenger loyalty and extract their information by a data mining technique from a database. Analytical results demonstrate that loyal passengers had high satisfaction in terms of service preferences, including airport service, passenger cabin facilities, information provision and complaint resolution, and flights departing on schedule. Loyal passengers also emphasized luggage services and obtaining airline information without an agency. The suggestions of this study not only provide Taiwanese airlines with a valuable reference for planning database marketing and managing loyal passengers but also expand the applicability of management information systems (MIS) to airline industry research.
**********
Air passenger numbers in Taiwan have been declining recently, from approximately 12 million in 2001 to around 8.6 million in 2006 (Taiwanese Ministry of Transportation and Communications, MOTC). Simultaneously, Taiwanese airlines have faced unprecedented competition with the launch of the Taiwan High Speed Rail in 2007, which is predicted to capture 50 percent of the passengers from the domestic air transportation market. Retaining passenger loyalty is thus crucial for Taiwanese domestic airlines.
Retaining passenger loyalty requires reducing or eliminating negative influences on service quality by ensuring customers high quality rate service (Gourdin and Kloppenborg 1991). Additionally, companies must closely monitor customer characteristics to accurately target desirable customers, considering not merely demographics but also consumption behaviors, preferences, etc. (Peppers, Rogers, and Doff 1999). Taiwanese airlines thus should effectively obtain loyal passenger decision information, including personal information, consumption behavior, and perceived service quality in order to retain passenger loyalty.
LITERATURE REVIEW
Previous decision information studies focused on air passengers generally applied a discrete choice model to understand passenger decision behaviors or preferences, including airline choice, airport choice, etc. (e.g., Pels, Nijkamp, and Rietveld 2001; Zhang and Xie 2005; Suzuki 2007). Recently a series of studies have attempted to expand the application of data mining (DM) techniques to deal with passenger decision information (Huehlin and Vannotti 2001; Giovanna and Silvia 2006; Wong, Chang, Jeng, Chung, and Lin 2006; Giuliano, Christian, and Andreas 2007) owing to recent rapid growth of databases. Numerous pattern recognition algorithms applied in DM have been developed to discover the complex relationships (e.g., non-linear) between input variables and output targets, which are difficult to identify using mathematical or statistical methods (Xie, Lu, and Parkany 2003). Furthermore, discrete choice model invariably appears to rest on the assumption of a utility-maximization framework derived from microeconomic theory (Ben-Akiva and Lerman 1985), and certain models (e.g., logit-based model) suffer from independence of irrelevant alternatives (IIA). It implies that adding another alternative or changing the characteristics of a third alternative does not affect the relative odds between the two alternatives. And it results in unrealistic applications with similar alternatives. Xie et al. (2003) pointed out that the linear property and synergy effects of the utility functions may not adequately modelize the comprehensive and complex correlations between independent and dependent variables. In contrast, DM techniques do not need statistical assumptions (e.g., no multi-collinearity) or other limitations (e.g., IIA property) between independent and dependent variables because they are based on an artificial intelligence perspective. Although numerous discrete choice models as well as DM technologies possess some similar advantages (e.g., non-linear modeling), DM techniques seem to reduce incompatibility between model structure and explanatory data.
Unlike most previous studies that employed DM techniques to handle aircraft repairing systems (e.g., Sylvain and Fazel 1999; Brence and Brown 2002; Knight, Cook, and Azzam 2005), data mining techniques now enable airlines to obtain a wealth of passenger information from their databases; for example, market segmentation and pricing (Toh and Raven 2003). This study focuses on loyal passenger management and develops a mining process using the DM technique to extract their decision information from the database of a Taiwanese airline. Additionally, this study compares the accuracy and intelligibility between a discrete choice model and a DM technique to identify whether the DM technique handles this issue more effectively.
Based on the research objectives, the following section addresses how to assess passenger loyalty and data mining context.
ASSESSMENT OF PASSENGER LOYALTY
The idea of "loyalty" was first introduced in the 1950s. Loyal customers give repeat business and promote the company through word of mouth (Dick and Basu 1994), leading to the establishment of a long-term relationship (Jones and Sasser 1995). Customer loyalty is thus fundamental to business profits. Customer loyalty is also viewed as the strength of the relationship between individual attitude towards an entity (brand, service, store, or vendor) and repeat patronage (Dick and Basu 1994). More recently, researchers have proposed various measuring indexes for assessing customer loyalty (Jones and Sasser 1995; Chaudhuri and Holbrook 2001), and these indexes can be divided into two types, namely behavioral and attitudinal. Behavioral assessment indexes include purchase number/quantity, purchase frequency, and period of ownership. Meanwhile, attitudinal assessment indexes include willingness to purchase again, willingness to recommend the product to others, public praise, self-loyalty, and recognition.
Behavioral indexes are more clearly observed than attitudinal indexes and more suitable for measuring passenger loyalty in airline databases because consumption records (e.g., flying frequency) in such databases imply passenger loyalty. However, the difference in assessing the validity by comparing behavioral and attitudinal aspects is only slight since buying actions are moderated by...
|
