| Summary: | Discrete choice models are a particular class of models which are applied when analysing a decision maker's choice from a set of alternatives. One of the most commonly applied models is the Nested Logit (NL) model, where the set of alternatives are separated into groups or nests, with no overlap. The Cross-Nested Logit (CNL) model allows for an overlap between nests, where each alternative belongs in part, to a particular nest. Alternatives which are grouped together are assumed to exhibit similar unobserved characteristics influencing each decision maker's choice. The analyst assigns each of the alternatives to a nest. This is known as the nesting structure. Given many alternatives, it becomes near impossible to analyse each one. There is also the additional complication that many of the nesting structures will output infeasible parameter estimates in terms of the nesting parameters. The approach developed in this thesis is autonomic in the sense that a nesting structure is empirically generated automatically, therefore removing the need for the analyst to impose a potentially infeasible nesting structure on the data. The Autonomic Nested Logit (ANL) and Autonomic Cross-Nested Logit {ACNL) models developed in this thesis, use a nesting structure that has been output from performing cluster analysis on the data. The number of clusters is determined using an observation weighted version of the cophenetic correlation coefficient. Given sufficient variables to segregate the data, these autonomic models show promising results. This autonomic approach is applied to data concerning a student's choice of Further Education campus in Northern Ireland in the academic year 2008/2009. The resulting model can act as a decision support tool to inform investment strategy regarding Further Education infrastructure in Northern Ireland.
|
|---|
