Motalleb Nejad, MohammadFlooding is a significant disaster that threatens the lives of millions of people within the United States. The 100-year coastal floodplain in the United States accommodates over 60,000 miles of major roadways. Moreover, United States Census Data reveals that more than 50% of the U.S. population lives close enough to the shoreline to be influenced by potential flooding. One major problem in this regard is identifying people who strongly depend on public transportation.A Conditional Wasserstein Generative Adversarial Network model and a couple of statistical approaches are introduced to capture the interdependency of people's characteristics in hazard zones. This method can provide a robust tool to estimate the demand for evacuation. An available statistical tool, namely the copula model, and its limitations are discussed, and the superiority of GANs model in addressing these limitations is mentioned. Also, a dynamic mixed integer linear problem has been developed to use the estimated demand and minimize the total evacuation time, the total departure time, and the number of utilized shelters. The model is solved for a simple network. A heuristic genetic algorithm is designed to solve the problems of actual life scales.enDissertationCivil engineeringTransportationConvolution Neural Nets and Computer VisionEvacuation Planning ModelsGenerative Adversarial Nets (GANs)Machine learning and Data ScienceNondominated Sorting Genetic AlgorithmSynthetic Data Generation