Research on Flight Dynamic Price Adjustment Based on Two-Step Passengers Choice Behavior Evaluation

Abstract

Dynamic pricing represents an emerging trend in the domain of research and innovation within airline revenue management. A crucial aspect of dynamic pricing involves the modeling of passenger choice behavior, particularly the challenge of estimating the behaviors of unobservable passengers who choose not to purchase. Traditional research predominantly employs the Multinomial Logit (MNL) model and leverages algorithms such as Expectation Maximization (EM) or Markov Chain Monte Carlo (MCMC) algorithm to address this challenge. Nonetheless, these algorithms are difficult to directly apply in practice due to long computation times and the scarcity of literature considering competitive factors. In this paper, we initially undertake the task of quantifying competitive factors and integrating these into the factors impacting passenger choice behavior. Subsequently, we introduce a two-step non-homogeneous estimation method that decomposes the log-likelihood function into marginal and conditional components to evaluate the parameters of passenger choice behavior. Employing this strategy allows us to obtain the probability of passenger purchases and passenger arrival rate across diverse pre-sale periods. In conclusion, this study introduces a sophisticated dynamic price adjustment model, meticulously designed for the segmentation of cabin classes based on logical criteria. The empirical validation of this model lends credence to its efficacy, showing that the passenger choice model considering competitive factors results in alignment with real-world flight sales situations, particularly in estimating passenger choice behavior and assessing arrival rates. Remarkably, when juxtaposed with prevailing airline pricing strategies, our proposed dynamic pricing adjustment strategy demonstrates a significant elevation in average flight revenue.