|Ph.D Student||Shmueli Deborah|
|Subject||Comparative Analysis of Travel Patterns of Men and Women|
|Department||Department of Architecture and Town Planning||Supervisor||Professor Emeritus Daniel Shefer|
This thesis examines the applicability of neural network modeling to the analysis of travel issues, and its extension to broader transportation planning. The research borrows a model developed in the field of Artificial Intelligence to study the complex relationships involved in transportation planning, and to test whether neural networks can be an effective predictive and explanatory methodology. Building on prior research, it analyzes and compares methodological approaches to test the advisability and practicality of adopting a method not widely used in the social sciences. This is the first time that this methodology has been applied to explore issues of trip generation and travel behavior.
The thesis compares the travel demand patterns of men and women, a subject which has layers of variables that include income, employment, residential location, household size and structure, and transportation resources, as well as a range of sociological and personal factors. The relevance of the topic is heightened by the steadily increasing percentage of women in the paid labor force, as well as trends toward suburbanization and the two—car family. These tendencies require consideration in future public policy decisions, which must address the symbiotic relationship between transportation and land use planning.
The information base used is the Traveling Habits Survey of 1984. This was combined with demographic and socio-economic data of the 1983 Population and Housing Census. The simulation tool used is Explorations in Parallel Distributed Processing, written in C.
The conclusions analyze two sets of hypotheses — for travel behavior and for neural networks. For travel behavior, the results confirm the basic tenet that gender is a determinant variable of travel, and support many of the other hypotheses about the differences in women’s travel patterns. In analyzing the neural network hypotheses, the conclusion is that the network, after learning, is an effective and useful predictive tool. When a functional relationship exists, it learns very well; when a functional relationship does not exist, it learns statistical correlations.