Dependent Demand Forecasting Models in Airline Revenue Management: Parametric Estimation Using Simulation

Joint forecasting models (JFMs) in airline revenue managementRevenue management are dependent on demand forecastingDemand forecasting models that predict demand volume and consumer choice behaviour simultaneously. JFMs consider various factors such as the booking curve, seasonality, “maximum-willingness-to-pay” (MWTP) of the customer, and attributes of the available products. JFMs are parametric models that use mixed logit modelsMixed Logit Model to study customer choice behaviour, where the estimation of model parameters is a challenging problem. We propose a sequential two-stage hybrid modellingHybrid modelling approach (a simulationSimulation-based heuristic algorithm) for parameter estimation of joint forecasting models using the Airline Planning and Operations Simulator (APOS) on actual airline data. In the first stage, parameters of demand volume are estimated by nonlinear least squares approximation using the Levenberg-Marquardt algorithm. In the second stage, MWTP parameters and choice-logit parameters are estimated simultaneously in two steps involving clustering and a simulationSimulation-based heuristic algorithm using Quai Monte-CarloMonte-Carlo SimulationSimulation. Forecast predictions obtained using this method provide accurate results with lesser computational effort when compared with the results obtained using complete simulationSimulation through pseudo-Monte-CarloMonte-Carlo methods. We recommend this algorithm for enabling faster computation in the estimation of parameters and optimal seat allocation and pricing in choice-based networks in airline revenue managementRevenue management.