IN-PERSON: Livable Cities – New York

Abstract

Performance benchmarking in urban planning is a method to evaluate how well a planned development will perform when built with respect to a set of key indicators by comparing it with another existing development. This study describes an innovative algorithm developed to evaluate the urban form of any existing or future development by benchmarking using K Nearest Neighborhood (KNN) machine learning. The methodology applied in this study can match a proposed Master Plan development with the closest existing development among the 220,740 US Census Block Groups (CBG). The Master Plan can then change the proposed characteristics related to urban form to match with another CBG until the client’s vision is satisfied. Many studies indicate that urban form indicators can evaluate livability. Selected indicators to represent urban form characteristics are the same as the variables applied in the National Walkability Index (NWI) created as part of the U.S. EPA Smart Growth Program. These include walkability, proximity to public transit, land use mix, and density. This study initially elaborates on the benefits of each indicator. Application of the same indicators using the NWI is then assessed. Finally, a machine learning methodology using K Nearest Neighborhood (KNN) is applied to predict the closest match of any development. The results help city planners evaluate the pros and cons of their plans if implemented and also guide any required modifications. The KNN machine learning code was created in Python as part of this study.