Segmentation of Airline Markets Using Descriptive and Cluster Analyses

Click to enlarge


Catherine Cleophas (RWTH Aachen University, Germany)


Sebastian Vock

Partner Institutions:

Airlines Reporting Corporation
Global Eagle Entertainment

In recent years, there has been interest in forecasting demand for revenue managment (RM) applications at the network (versus flight) level. That is, instead of forecasting demand and optimizing the set of offered products for a single flight, researchers forecast demand and control product availablity for directional itineraries. Origin-destination pairs (OD) consisting of at least one flight and several additional characteristics, such as point-of-sale, departure date, and departure time define these itineraries. The transition to network models has led to many successful state-of-the-art RM methods (e.g., see [1,2,3]).  However, using this high dimensional information increases the complexity of the RM model and introduces the problem of small numbers, counteracting the achieved benefits of the model [4]. Therefore, it is beneficial to aggregate itineraries to achieve a reasonable level of detail and preserve practical applicability.

The majority of current aggregation methods for RM applications are based on geoographical charcteristics (e.g., grouping of OD pairs that originate in the U.S. and terminate in Europe) and expert opinion.  Air travel demand, however, is highly dependent on the characteristics of the specific itineraries. Therefore, aggregation models should use more detailed information and manifold dimensions to help ensure more homogenous groupings of markets.  Data mininig methods, which have been successfully applied in marketing, business analytics, and other fields are powerful tools for this purpose. Data mining involves the analysis of large data sets with the aim to gain knowledge about the data stored. Beside the extraction of information, data mining seeks to transform this information  into understandable structures that may be used for visualization and further analyses.

This research project examines the effect and success of data mining techniques to airline ticketing data in order to explore possible market segmentations. For this, graduate researcher Sebastian Vock already extracted a data set based on a ticketing database provided by the Airlines Reporting Corporation and scheduling database provided by Global Eagle Entertainment of approximately 23,000 itineraries from eight different U.S. airlines to study the success of different pre-processing methods. Based on the distribution of bookings over a specific fare range, the next steps should lead to insights about market structures included in this data set. To this end different cluster analyses and other classification methods will be applied to the ticketing data. The objective is to find possibilities to classify itineraries based on the underlying supply and demand structure by using data mining technologies and without pre-defined classification schemes. A respective classification may be an opportunity to aggregate current data before applying RM methods in order to reduce the drawbacks of huge complexity and small numbers.



[1] Vock, S., Garrow, L.A. and Cleophas, C. New perspectives on airline ticket data – data-driven approaches to cluster air travel itineraries.  (Working paper under review).


[1] Bertsimas, D. & Popescu, I., 2003. Revenue management in a dynamic network environment. Transportation Science, 37(3), pp.257–277.

[2] Curry, R.E., 1990. Optimal airline seat allocation with fare classes nested by origins and destinations. Transportation Science, 24(3), pp.193–204.

[3] Talluri, K. & van Ryzin, G., 1998. An analysis of bid-price controls for network revenue management. Management Science, pp.1577–1593.

[4] Bartke, P., Cleophas, C. & Zimmermann, B., 2013. Complexity in airline revenue management. Journal of Revenue & Pricing Management, 12(1), pp.36–45.