Deterministic mathematical programming models that capture network effects play a predominant role in the theory and practice of airline revenue management. These models do not address important issues like demand uncertainty, nesting, and the dynamic nature of the booking process. Alternatively, the network problem can be broken down into leg-based problems for which there are satisfactory solution methods, but this approach cannot be expected to capture all relevant network aspects. In this paper, we propose a new algorithm that addresses these issues. Starting with any nested booking-limit policy, we combine a stochastic gradient algorithm and approximate dynamic programming ideas to improve the initial booking limits. Preliminary simulation experiments suggest that the proposed algorithm can lead to practically significant revenue enhancements.
This thesis studies a well-known problem in the airline industry, namely, the seat-allocation problem in a single leg of an airline flight. The problem studied considers random arrivals, cancellations, no-shows, multiple fare classes, and overbooking. A simulation-based approach using simultaneous perturbation and simulated annealing was used to solve the problem. The performance of these techniques was benchmarked with a widely used heuristic, namely, expected marginal seat revenue heuristic. The results obtained show that the simulation-based optimization approach can outperform the heuristic approach.
Accurate forecasts are crucial to a revenue management system. Poor estimates of demand lead to inadequate inventory controls and sub-optimal revenue performance. Forecasting for airline revenue management systems is inherently difficult. Competitive actions, seasonal factors, the economic environment, and constant fare changes are a few of the hurdles that must be overcome. In addition, the fact that most of the historical demand data is censored further complicates the problem. This dissertation examines the challenge of forecasting for an airline revenue management system in the presence of censored demand data. This dissertation analyzed the improvement in forecast accuracy that results from estimating demand by unconstraining the censored data. Little research has been done on unconstraining censored data for revenue management systems. Airlines tend to either ignore the problem or use very simple ad hoc methods to deal with it. A literature review explores the current methods for unconstraining censored data. Also, practices borrowed from areas outside of revenue management are adapted to this application. For example, the Expectation-Maximization (EM) and other imputation methods were investigated. These methods are evaluated and tested using simulation and actual airline data. An extension to the EM algorithm that results in a 41% improvement in forecast accuracy is presented.
Revenue Management (RM) in the airline industry plays a very important role in maximizing revenue under various uncertainty issues, like customer demand, the number of seats to be maintained in inventory, the number of seats to be overbooked, etc. In this dissertation, a Markov decision process (MDP) based approach using statistical modeling is presented. Prior versions of this statistical modeling approach have employed remaining seat capacity ranges from zero to the capacity of the aircraft. In reality, actual remaining capacities are near capacity when the booking process begins and near zero when the flights depart. Thus, our modified version uses realistic ranges to enable a more accurate statistical model, leading to a better RM policy. We also consider overbooking, no-shows and cancellations and estimate the optimal number of seats to be overbooked using a hybrid approach that combines Newton's and steepest ascent method. The extended statistical modeling approach in this dissertation consists of three modules: (1) the revised statistical modeling module, (2) the overbooking module, and (3) the availability processor module. The first two modules are conducted off-line to identify optimal overbooking pads and derive a policy for accepting/rejecting customer booking requests. The last module occurs on-line to conduct the actual decisions as the booking requests arrive. To enable a computationally-tractable solution method, the revised statistical modeling module, under an assumed maximum overbooking pad of 20%, consists of three components: (1) simulation of the deterministic bid price approach to identify the realistic ranges of remaining seat capacity at different points in time; (2) solutions to deterministic and stochastic linear programming problems that provide upper and lower bounds, respectively, on the MDP value function; and (3) estimation of the upper and lower bound value functions using statistical modeling. Next, the overbooking module identifies the optimal number of seats to be overbooked. Finally, the value function approximations are used with the optimal overbooking pads to determine the RM accept/reject policy.
Revenue management (RM) has emerged as one of the most important new business practices in recent times. This book is the first comprehensive reference book to be published in the field of RM. It unifies the field, drawing from industry sources as well as relevant research from disparate disciplines, as well as documenting industry practices and implementation details. Successful hardcover version published in April 2004.
The decision to buy a fare that is higher than the desired fare, under the situation when the desired fare is not available, is known as sell-up. Passengers' willingness to sellup can have considerable impact on airline revenues. The extent of this impact is dependent upon the method used to control booking limits and other parameters associated with passenger demand and fare structure. In this thesis we demonstrate the importance of incorporating sell-up in airline revenue management. The improvement in revenue, under various scenarios, and under various seat inventory control algorithms, is discussed. We also analyze the influence of demand factor, spill, sell-up rate and fare ratio on these improvements. A modification of the EMSRb heuristic is proposed to capture the revenue potential associated with passenger sell-up. The proposed rule increases the protection levels, obtained from the EMSRb algorithm, as long as the expected gain, from every additional seat protected, is greater than the expected loss. Unlike the existing models, the proposed rule uses expected spill to determine the expected number of passengers that would sell-up at a given demand level and sell-up rate, and then adjusts the protection levels accordingly. This makes it robust to variations in demand levels. We have also developed a simulation to compare the performance of the existing rules with that of the proposed heuristic. The simulation has the ability to account for errors in sell-up estimation and variability in demands. It is shown that the revenue gains under the proposed rule may not exist under all situations. In the tests performed in this thesis, the improvements over the original EMSRb algorithm vary from 0% to over 2.5%. Although the gains are not consistent, the proposed rule does not cause any negative impact on overall revenues and hence is unlikely to pose any risk when implemented over the original EMSRb algorithm.
Simulation-Based Optimization: Parametric Optimization Techniques and Reinforcement Learning introduce the evolving area of static and dynamic simulation-based optimization. Covered in detail are model-free optimization techniques – especially designed for those discrete-event, stochastic systems which can be simulated but whose analytical models are difficult to find in closed mathematical forms. Key features of this revised and improved Second Edition include: · Extensive coverage, via step-by-step recipes, of powerful new algorithms for static simulation optimization, including simultaneous perturbation, backtracking adaptive search and nested partitions, in addition to traditional methods, such as response surfaces, Nelder-Mead search and meta-heuristics (simulated annealing, tabu search, and genetic algorithms) · Detailed coverage of the Bellman equation framework for Markov Decision Processes (MDPs), along with dynamic programming (value and policy iteration) for discounted, average, and total reward performance metrics · An in-depth consideration of dynamic simulation optimization via temporal differences and Reinforcement Learning: Q-Learning, SARSA, and R-SMART algorithms, and policy search, via API, Q-P-Learning, actor-critics, and learning automata · A special examination of neural-network-based function approximation for Reinforcement Learning, semi-Markov decision processes (SMDPs), finite-horizon problems, two time scales, case studies for industrial tasks, computer codes (placed online) and convergence proofs, via Banach fixed point theory and Ordinary Differential Equations Themed around three areas in separate sets of chapters – Static Simulation Optimization, Reinforcement Learning and Convergence Analysis – this book is written for researchers and students in the fields of engineering (industrial, systems, electrical and computer), operations research, computer science and applied mathematics.
Erfolgsmessung ist ein integraler Bestandteil jedes Revenue Management (RM)-Systems. Das Revenue Opportunity Model (ROM) ist ein bekanntes Verfahren zur Erfolgsmessung. Bei der Anpassung des ROM an die wesentlichen Entwicklungen des RM insb. die Entwicklung von Einzelflug-basierter zu Netzwerk-basierter RM Steuerung und der Übergang von unabhängiger zu abhängiger Nachfrage - gewinnt die Frage der Anwendbarkeit und insb. der Validität des ROM an Bedeutung. Wir modellieren unabhängige und abhängige Nachfrage in einem Netzwerk-basierten ROM und untersuchen seine wichtigsten Eigenschaften. Darüber hinaus betrachten wir verschiedene praktische Aspekte des RM einer Netzwerk-Fluggesellschaft und untersuchen die Aufsplittung der aggregierten Kennzahlen des Netzwerk-basierten ROM auf einzelne Flüge und auf einzelne Komponenten des RM. Wir stellen einen neuartigen simulationsbasierten Ansatz zur Untersuchung des ROM vor insb. der Robustheit gegenüber Fehlern in der unbeschränkten Nachfrageschätzung. Das in der Simulationsumgebung verwendete RM-System spiegelt alle wesentlichen Komponenten und RM-Methoden wider, wie sie von einer Netzwerk-Fluggesellschaft verwendet werden, insb. hinsichtlich Prognose- und Optimierungsmodellen. Bei realistischen Eingabedaten erweist sich das Netzwerk-basierte ROM, sowohl mit unabhängiger als auch mit abhängiger Nachfrage, als robust gegenüber Fehlern in der unbeschränkten Nachfrageschätzung. Darüber hinaus ist eine Aufsplittung der ROM Kennzahlen auf einzelne Flüge möglich. Allerdings zeigt sich das ROM auf Einzelflugebene weniger robust. Weiterhin wird ein Ansatz vorgestellt, der den Beitrag von Überbuchung, Upgrading und Tarif-Mix aus dem Gesamterfolg isoliert. Insbesondere erweisen sich die ROM Kennzahlen für Überbuchungen und Upgrading als sehr robust.. - Techniques for performance measurement are an integral part of a revenue management (RM) system. The Revenue Opportunity Model (ROM) is a widely known method to measure revenue management performance. While adapting the ROM to recent developments of revenue management science - i.e. the advancement from leg-based to network-based RM controls and the recent transition from independent to dependent demand structures - the question of applicability and in particular the validity of the ROM have become increasingly important. In this thesis we model both independent and dependent demand structures in a network-based ROM and investigate its main properties. Furthermore, we consider different aspects of airline RM to make the application of the ROM possible in practice. We not only cover a disaggregation of the aggregated network ROM measures to single legs, but also to single RM components. In this thesis we therefore introduce a novel simulation-based approach to investigate the ROM properties particularly to measure its robustness against errors in the estimated unconstrained demand. The RM system used in the simulation environment reflects all main components and RM methods in use of a large network airline, particularly regarding state-of-the-art demand modeling and optimization models. The network-based ROM, both with independent and dependent demand, proves itself robust against errors in the estimated unconstrained demand for realistic input data with better robustness results for the ROM with independent demand. Moreover, a disaggregation of the ROM measures to single flight legs is possible. However, the ROM shows less robustness on a leg level. In addition, an approach to isolate the contribution of overbooking, upgrading and fare-mix from the overall success is introduced. In particular, the ROM measures for overbooking and upgrading prove themselves to be very robust.
