Project by Daegeun Kim
This project presents a quantitative geometric analysis of street blocks across the top ten U.S. metropolitan areas by GDP: New York, Los Angeles, Chicago, Dallas, Washington D.C., San Francisco Bay Area, Boston, Houston, Atlanta, and Seattle. Focusing purely on physical geometry, four shape-based metrics: Area, Isoperimetric Quotient, Rectangularity, and Solidity were derived from each block to characterize its form and compactness. A exponential regression model was developed to evaluate which of these four geometric indicators best predicts population density, and to assess the independent usefulness of each metric as a predictor. In parallel, a logistic regression model was trained to classify the urbanity of each block using the dataset’s original UR20 label, which defines urban versus non-urban areas based on criteria such as population concentration, adjacency of developed blocks, and built environment continuity. Model reliability and generalization were validated through 10-fold cross-validation, ensuring robustness against sample bias. This study aims to isolate geometric form as an explanatory variable in urban analysis, demonstrating that the morphology of street blocks alone can provide meaningful insight into patterns of density and urbanization across major U.S. cities.
