Friday 28 March 2025
Urban planning is a complex and multifaceted discipline that has been studied for centuries. From the earliest urban settlements to modern-day metropolises, cities have always been hubs of human activity, commerce, and innovation. Yet, despite their importance, cities are often plagued by problems such as overcrowding, pollution, and social inequality.
In recent years, researchers have turned to data analysis and machine learning techniques to better understand the intricate relationships between urban form and function. One such approach is the use of graph neural networks (GNNs), which can learn patterns and relationships within complex systems like cities.
A team of researchers has developed a novel framework called CoMo, which uses GNNs to analyze the relationship between urban morphology and function. Urban morphology refers to the physical characteristics of buildings, streets, and other urban features, while function refers to the activities and purposes that these spaces serve.
The researchers used Boston as their case study, analyzing data on building footprints, street networks, and land use patterns. They found that by using GNNs to analyze this data, they could identify patterns and relationships that were not apparent through traditional methods.
For example, the team discovered that buildings in different neighborhoods had distinct morphological characteristics, such as shape, size, and orientation. These characteristics were closely tied to the functions of the buildings, with residential areas featuring more compact and uniform buildings, while commercial districts had taller and more elongated structures.
The researchers also found that the relationships between urban form and function varied depending on the scale at which they were studied. At a small scale, individual building characteristics were most important, while at a larger scale, the layout of streets and neighborhoods became more significant.
One of the key advantages of CoMo is its ability to provide insights into complex urban systems in an interpretable way. Unlike traditional machine learning models, GNNs can be trained to produce visualizations that reveal the underlying patterns and relationships within the data.
The researchers hope that CoMo will have practical applications for urban planners and policymakers, who can use the framework to design more effective and sustainable cities. By better understanding the intricate relationships between urban form and function, cities can become more livable, equitable, and resilient.
In addition to its potential practical applications, CoMo also highlights the importance of interdisciplinary research in urban planning. By combining insights from geography, architecture, sociology, and computer science, researchers can develop a more comprehensive understanding of complex urban systems.
Cite this article: “Unraveling Urban Complexity: A Novel Framework for Analyzing City Form and Function”, The Science Archive, 2025.
Urban Planning, Machine Learning, Graph Neural Networks, Como, Urban Morphology, Function, Boston, Building Footprints, Land Use Patterns, Sustainability
