New Heuristics for Revenue Management Problem with Customer Choice Models.

This dissertation focuses on designing efficient heuristics for revenue management problem with customer choice behavior. This problem has gained growing interest in recent years as a result of fierce competition, fare transparency and restriction free pricing. Existing solutions to RM with restriction free pricing are either based on ad-hoc adjustments to traditional capacity allocation or computationally intensive dynamic programming. In our framework, demand for different fare classes is estimated by a customer choice model. We show how a multinomial logit demand model can be used to provide upsell estimates for handling dependent demands and also account for competitive effects. In chapter 2, we present new EMSR-based formulations for the single-leg, nested, multiple fare RM problem that can be used to solve both dependent and independent demand problem. We develop efficient and nearly optimal static heuristics for RM optimization that are more general and provide better performance than the widely used EMSR-b algorithm for independent demands. Variations of the algorithm for both low-to-high and mixed arrival order cases are included. Chapter 3 deals with the dynamic formulation for the single-leg problem (Talluri and van Ryzin (2004a)) where the control policy depends on remaining time and remaining capacity. We develop near optimal and computationally efficient heuristics which generate control policy without having to solve the computationally intensive dynamic programming. The heuristics are designed for the case where the efficient sets are nested, as in the Multinomial Logit Model or the independent demand model. The nesting, however, does not necessarily need to be by fare values. The extensions to RM problem with time heterogenous choice models as well as non-homogenous customer arrivals are also studied. The numerical experiments indicate that the heuristics perform well. Parallel flights problem with two fare classes is studied in Chapter 4. We assume that low fare demands come first in stage 1 followed by high fare demands in stage 2. We consider simultaneous seat inventory control of low fare class on these n parallel flights where the spilled demand can either upsell to the high fare on the same flight or goes to alternative same fare class on other flights. We provide both the upper bound and the lower bound on the value function and propose a heuristic which is based on capacity pooling technique together with the static Choice-Based EMSR heuristic. The conclusions are summarized in the last chapter.