Network Migration Strategies: Evaluating Performance with Extensions of Data Envelopment Analysis.

Abstract

In the current economic climate, many companies are facing pressures to operate more efficiently resulting in frequent mergers, acquisitions, and reorganizations. This often necessitates a change in various network structures within these organizations, which leads to the guiding question of this study. How does an organization efficiently and effectively transition its network structures while making use of multiple performance measures? This dissertation seeks to develop models and algorithms, based on Data Envelopment Analysis (DEA), to analyze perturbations in real-world network topologies. The first part of this dissertation shows the historical development of DEA with an introduction of notation and extensions. The next chapter is a detailed study of US airport inefficiency, which answers the question: Is there a difference in the efficiency of hub and non-hub airports in the United States? After decomposing the efficiency into scale efficiency, mixed efficiency and pure technical efficiency, the Wilcoxon Rank Sum test is able to show that there are differences between hub and non-hub airports. The fourth section of the dissertation develops a theoretical model, the Inverse Range-based Directional Distance model (INVRDD-DEA), to address the presence of reverse quantities in DEA and yield shortest path projections. This model leads to the Fully Comprehensive RDD-DEA model that takes into account all sources of inefficiency. This model is then used to illustrate all sources of inefficiency in a greenhouse gas example. The final contribution of the dissertation is an exploration into the evolution of the operations research approach to the field of network science. The concept of re-engineering of networks, defined as the ability to optimize perturbations to existing networks based on several performance metrics, is used as a methodological model for typical types of changes that exist in cooperate networks. The critical factors for building an algorithm for modifying network topologies are identified and used to design a procedure for making changes in networks. Finally, an example of an ERP implementation is given to show the benefits of using DEA to make changes to existing network topologies.