"A Self-Organizing Neural System For Urban Design"
The dynamics of urban systems are characterized by complex non-linear relationships between socio-economic attributes of land use and spatial interactions. Traditional urban models have many limitations in simulating these urban dynamics. To overcome these limitations, a number of new approaches have been adopted. This paper examines a neural network based approach to the analysis of growth factors in an urban design proposal. The system incorporates Kohonen’s self-organizing map algorithms within an existing GIS application to function as a design and decision support system. Urban data of a simulated region is embedded in the neural net and correlated, in varying degrees, with data obtained from case studies and/or other local regions. This allows the user to visualize and understand the impacts of the proposal, which is otherwise difficult to envision because of its complexity.