The proliferation of Airbnb was found to be associated with negative externalities, such as rising rents, house prices and gentrification. Nonetheless, it also seems to capture “positive” urban qualities related to the ordinary/traditional built environment. However, this relationship is still largely understudied as most works focus on small samples (a specific neighbourhood) and limited numbers of urban descriptors.
In this work, we used the Urban MorphoMetric (UMM) approach in combination with machine learning techniques to investigate the relationship between urban form and density of Airbnbs in the whole city of Amsterdam (NL). More specifically, we, first, identify urban types (UTs), i.e. areas with homogenous patterns of urban form, via the clustering of a comprehensive set of 296 morphometric descriptors. Second, we investigate the relationship between UTs and density of Airbnbs via a composite machine learning technique, based on forward feature selection, gradient boosting and an interpretative tool of machine learning outputs. Third, we offer visual and textual profiles of the UTs most positively and negatively associated with density of Airbnbs.

Figure 2. Morphometric taxonomy of Kochi. Buildings are color-coded according to their respective UTs and level of similarity.
Results show that 15 UTs on 21 explain up to 44% of Airbnb density’s variance. UTs characterised by a compact and more diverse urban fabrics are positively associated with Airbnbs. These largely correspond to the Oud-West, Oosteljke Eilanden and Weesperburt en Plantage areas. Conversely, UTs featuring a more repetitive and sparse urban fabric are inversely associated with Airbnbs. These UTs mainly correspond to areas located near the A10 ring road and the western part of Amsterdam-Zuidoost.
This work provides a novel and replicable way to investigate the intricate relation between urban form and Airbnb at an unprecedented scale. Furthemore, by identifying features related to urban attractiveness, it can potentially constitute the basis for the generation of evidence-based urban design codes, incorporating sought-after and place-making qualities, in existing and new neighbourhoods.

Figure 3. Sample sites with existing fabric (top left) and three different figure-ground design demonstrations (top right, bottom left, bottom right) proposed by three different designers, in compliance with the morphometric profile of UT1.
Recent posts:
-
Urban form and COVID-19 in Greater London: an urban morphometric approachJune 16, 2022/0 Comments
-
Archive:
- June 2022 (1)
- April 2022 (2)