To lay the foundation for this work’s motivation in Chapter 2, this chapter provides a quick overview to the specifics of the passenger airline industry in general and details two recent, significant changes in its structure and conduct in particular.
Following the introduction to the relevant underlying subjects of this work that has been given in the previous chapter, the particular motivation for selecting the overall topic and taking the specific approach are discussed here. First, Section 2.1 addresses the question of the topic’s overall relevance. This is followed by a justification of the specific focus on the airline industry in Section 2.2. Finally, Section 2.3 details the particular objectives and differentiating elements of this work before its structure is summarized in Section 2.4.
The present chapter’s objective is threefold. First, it locates advanced pricing and revenue optimization schemes within the general pricing context in order to highlight their application areas and principal characteristics (Section 3.1). Second, it gives an overview of prevalent literature and optimization models within that context (Section 3.2). Third, it discusses the shortcomings and limitations of dynamic pricing models developed throughout the literature (Section 3.3) to finally highlight a suitable scope for the work at hand (Section 3.4).
This chapter covers the theoretical and technical background of the specific learning method later employed in Chapters 6 and 7 to actually forecast latent demand based on its characteristics as described in Chapter 5. The presented method rests on the Bayesian interpretation of probability, which is fundamentally different from the classical or frequentist interpretation, where probabilities are simply viewed “in terms of the frequencies of random, repeatable events” (see, e.g., Bishop, 2006, p. 21).
This chapter describes the structure and characteristics of the collected demand data, which form the basis for the development of the corresponding forecasting model in Chapter 6.
This chapter’s objective is to develop an overarching linear basis function model to forecast demand, which is done in Section 6.1. To ensure that it includes all relevant demand drivers, the model is validated following a typical frequentist interpretation and using the appropriate tests in Section 6.2 before the approach is finally extended to allow for Bayesian learning in Section 6.3.
Following the introduction of the Bayesian self-learning forecasting scheme underlying this work (see previous Chapters 4 – 6), this chapter now provides the computational results and takes a look at the overall predictive performance of the model in Section 7.1 as well as its sensitivity to using informative priors, changing learning window sizes and different forecast granularities in Section 7.2.
Preceeding Chapters 4 – 7 developed, validated and finally evaluated a Bayesian forecast model for latent demand in low-cost air travel markets; that is, uncensored demand – unbiased by pricing decisions – is forecasted directly online from deterministic input data.
The following chapter introduces the methodology used in this part of the work to formulate a model of the price sensitivity of customers within the low-cost travel market that is based on real-world data. Following many other works on choice analysis (see below Table 9.1), the method of discrete customer choice analysis is employed here to model and understand customer purchasing behavior at a disaggregated individual decision maker level.
This chapter presents and descriptively analyzes a proprietary dataset, which contains information on the choice situation of air travel customers in lowcost markets. The analysis is based on the structure and characteristics of exclusively collected real-world data that later also form the basis for the development of a corresponding discrete choice model in Chapter 11.
he objective of this chapter is to develop a multinomial logit model, as introduced in Chapter 9. The model simulates low-cost travel choice based on the market’s demand specifics as identified in Chapter 10.
wo statistically sound choice models have been developed in Chapter 11. This chapter examines the overall predictive performance of these models in the first Section 12.1. Additionally, specific elasticities and substitutional patterns are inferred in Section 12.2, and finally Section 12.3 draws targeted conclusions on the usage of the reported results in actual airfare pricing.
The preceding Chapters 9 – 12 developed, validated and finally evaluated a multinomial logit model to understand the particular customer choice probabilities that convert latent demand in low-cost air travel markets to eventual realized demand, based on the prevalent price environment in the market and the decision makers’ determining characteristics.