City Rhythm Data Model Review
Part of
by Julia Ubeda
The City Rhythm Data Model was developed by the "City Rhythms" research team in 2017. The published exploration includes a general description of the data model and the main outputs (see the links below the body text if you want to read the publication).
After the validation session that concluded the 2017 exploratory research project, the research team agreed that the development of the data model and the outputs had to be better understood since certain concepts in the data model were not defined in detail.
The research team asked the TU Delft PhD candidate, Julia Ubeda, to review the data model and to gather the technical details about it when she joined the "Designing Rhythms for Social Resilience" (2019-2023) new team.
Julia interviewed Scott Cunningham (TU Delft), who was the main Data Model researcher from the City Rhythms team, Eric Boertjes (Blooming data / AMS) and Ruben Spruit (Delph BI). Julia also consulted Caroline Nevejan, principal investigator of the "City Rhythms" and "Designing Rhythms for Social Resilience" projects and Pinar Sefkatli, UvA PhD candidate in order to get a better understanding of the working process within the group.
The following document comprises and summarises the insights that Julia got from those conversations and the data exploration.
The Base Rhythms concept was one of the main model outputs of the "City Rhythms" exploration and it is a very strong concept. The project stakeholders also agreed that, if they were aware of the base characteristics of a neighbourhood, it would be simpler for them to make policy-decisions and design interventions based on what's needed. Therefore, the new "Designing Rhythms for Social Resilience" (DRSR) research group would like to keep building on the idea of Base Rhythms to cluster social dynamics in neighbourhoods.
However, there are some drawbacks of using the same modelling approach and the same procedures in the future. The drawbacks and limitation are explained in the presentation and summarised in the next points:
- The model (the Markov chain) did not considered temporal variations from cluster to cluster (from Base Rhythm to Base Rhythm). It only looked into one-year transition and it forgets from which state the grid cell is coming from. The cell will remain in the main cluster during the four years time span that the project considered.
- A complete new variable had to be introduced [model requirement] that it is not related to the CBS datasets' variables. That hidden variable is what it is considered as the latent part of the model. Delph Bi decided to consider as that variable the distance between the CBS grids. The reasoning behind this assumption is that cell grids that are close to each other would probably have similar dynamics. However, the impact of this assumption on the model outputs was not tested. It is also difficult to observe a clear pattern based on the visualisations from AHTI and Eric.
- An idea about this hidden variable would have been introducing more information about social dynamics: how is your area developing in the last fifteen years.
- We do not know what the different clusters (Base Rhythms) mean. Since this is an unsupervised model, there is no way to give an interpretation to the model output and it is not possible to see what each base rhythm is composed of.
Nevertheless, The DSRS research team is willing to keep building on the Base Rhythms concept and they will explore and experiment with other modelling approaches together with the new project partner, Habidatum.
-------------------------------------------------------------------------------------------------------------------------------------------
The exploration research project is published here: Nevejan, C., Sefkatli, P. and Cunningham, S. City Rhythm : Logbook of an Exploration. Delft University of Technology. 2018 ISBN 978 90 819839 1 4
An animation of the model outputs can be found here.