Award Date

August 2017

Degree Type

Thesis

Degree Name

Master of Science in Engineering (MSE)

Department

Civil and Environmental Engineering and Construction

First Committee Member

Alexander Paz

Second Committee Member

Mohamed Kaseko

Third Committee Member

Dave James

Fourth Committee Member

Brendan Morris

Fifth Committee Member

Justin Zhan

Number of Pages

55

Abstract

Mixed logit models are a widely-used tool for studying discrete outcome problems. Modeling development entails answering three important questions that highly affect the quality of the specification: (i) what variables are considered in the analysis? (ii) what are going to be the coefficients for these variables? and (iii) what density function these coefficients will follow? The literature provides guidance; however, a strong statistical background and an ad hoc search process are required to obtain the best model specification. Knowledge of the problem context and data is required. Given a dataset including discrete outcomes and associated characteristics the problem to be addressed in this thesis is to investigate to what extend a relatively simple metaheuristic such as Simulated Annealing, can determine the best model specification for a mixed logit model and answer the above questions. A mathematical programing formulation is proposed and simulated annealing is implemented to find solutions for the proposed formulation. Three experiments were performed to test the effectiveness of the proposed algorithm. A comparison with existing model specifications for the same datasets was performed. The results suggest that the proposed algorithm is able to find an adequate model specification in terms of goodness of fit thereby reducing involvement of the analyst.

Keywords

Discrete Outcome; Mixed Logit; Optimization

Disciplines

Economics | Engineering

Language

English


Share

COinS